This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Cubes and Boxes have Rupert’s passages in every direction111This article has been accepted for publication in American Mathematical Monthly, published by Taylor & Francis.

András Bezdek222Supported by NKFIH grant KKP-133864.    Zhenyue Guan    Mihály Hujter    Antal Joós
Abstract

It is a 300300 year old counterintuitive observation of Prince Rupert of Rhine that in cube a straight tunnel can be cut, through which a second congruent cube can be passed. Hundred years later P. Nieuwland generalized Rupert’s problem and asked for the largest aspect ratio so that a larger homothetic copy of the same body can be passed. We show that cubes and in fact all rectangular boxes have Rupert’s passages in every direction, which is not parallel to the faces. In case of the cube it was assumed without proof that the solution of the Nieuwland’s problem is a tunnel perpendicular to the largest square contained by the cube. We prove that this unwarranted assumption is correct not only for the cube, but also for all other rectangular boxes.

1 Rupert’s passages and Nieuwland’s constants

It is commonly agreed that Prince Rupert of the Rhine (1619-1682) was the first person, who posed the following puzzle.:

Rupert’s passage problem: Cut a hole in a cube, through which another cube of the same size shall be able to pass.

Refer to caption
Figure 1: The cube is Rupert. a) is a visual proof from 1816 [12]; b) is a proof without words from today.

The affirmative answer of Rupert’s passage problem is attributed to P. Nieuwland and was published in a paper of J.H. van Swinden [12] in 1816. It is common to say that the cube is Rupert (see the survey papers [9] and [6]). On Figure 1b the four points marked by dots partition their edges in the ratio 1:31:3. Repeated use of Pythagorean theorem reveals that these four points form a square of edge length 324=1.06\frac{3\sqrt{2}}{4}=1.06\dots allowing to cut a suitable perpendicular passage [2].

Rupert’s passage problem sounds paradoxial at first. After realizing that the problem is about finding two shadows so that one shadow fits inside the other, the problem becomes manageable and also triggers generalizations. Which solids are Rupert? If a solid is Rupert, what is the maximum size of the second body so that it still can be moved through the first one? The second question was asked for cubes by P. Nieuwland (1764-1794) almost a century after Rupert’s question was posed. The later problem is referred to as Nieuwland’s passage problem and the homothetic constant corresponding to the largest size (or rather the supremum of the possible sizes) is called Nieuwland’s constant. Several survey papers popularized generalizations over the years, including papers of J.H. van Swinden [12] in 1816, of A. Ehrenfeucht [3] in 1964, of D.J.E. Schrek [10] in 1950, of M. Gardner [2] in 2001 and of V.F. Rickey [9] in 2005.

Over the years various solids were studied concerning Rupert’s passage problem. Spheres and bodies of constant widths are obviously not Rupert. Positive answers were found for tetrahedron and for octahedron by Scriba [11] in 1968, for rectangular boxes by Jerrard and Wetzl [6] in 2004, for universal stoppers by Jerrard and Wetzl in [5] in 2008, for dodecahedron and for icosahedron by Jerrard, Wetzl and Yuan [7] in 2017, for eight Archimedean solids by Chai, Yuan and Zamfirescu [1] in 2018 and for nn-cubes by Huber, Schultz and Wetzl [4] in 2018. The most intriguing open question in this area is wether every convex polyhedron is Rupert (see [7]).

2 New results concerning Rupert’s passages

Neither Rupert nor Nieuwland, were precise about how they want to pass a cube through the other one. According to standards at their time, using words like ’passage’ and ’tunnel’ was enough to indicate that motion is expected to be a translation. There is another assumption which requires rigorous proof. We noticed this detail in a paper of Jerrard and Wetzel who wrote several interesting papers on Rupert’s passages. [6] appeared in this Monthly, and is about the solution of Nieuwland’s type problem for passing rectangular boxes through a unit cube. In the introduction the authors state a comonly used unwarranted assumption

Nieuwland’s passage problem of finding the largest cube that can pass through a unit cube is equivalent to finding the largest square that fits in the unit cube, because once the largest square is located, the hole through the cube having that largest square as its cross section clearly provides the desired passage.

Indeed, once the largest homothetic square is located, the hole raised perpendicularly over this square clearly provides the desired passage. The other direction of the equivalency statement can not be assumed without proof. It is very unlikely that Rupert and Nieuwland wanted to restrict passages to tunnels raised perpendicularly over squares contained in the unit cube. Thus, before this equivalency issue is settled the constant found by Nieuwland or later by other authors are only lower bounds for the Nieuwland’s constants of the cube and of other solids.

The main goals of this paper were i) to show that cubes are Rupert in every direction and ii) to show equivalency of the later two problems for cubes. Along the way we noticed that essentially the same arguments, with some modifications, prove the analogous theorems for all rectangular boxes. We will prove

Theorem 1.

The interior of every hexagonal projection of a rectangular box with sides abca\leq b\leq c contains a rectangle with sides aa and bb. Equivalently, rectangular boxes have Rupert’s passages in every direction not parallel to faces. In particular, cubes have Rupert’s passages in every direction not parallel to faces.

Using Theorem 1 we also prove

Theorem 2.

Let BB be a rectangular box with edge lengths abca\leq b\leq c. If a homothetic box λB\lambda B (λ>0\lambda>0) can be passed through BB by translation, then the interior of box BB contains rectangle with sides λa\lambda a and λb\lambda b. With other words, the problem of finding Nieuwland constant of a given rectangular box BB, is equivalent to finding the supremum of λ\lambda’s for which the interior of box BB contains a λ\lambda-homothetic copy of the smallest face of box BB.

3 Preliminary Lemmas

In this section we prove three lemmas:

Lemma 1.

Let CC be a unit cube in a general position in the xyzxyz-coordinate space. Denote by HH the perpendicular projection of CC on the xyxy-plane (Figure 2). Then,

  1. 1.

    the projection HH is either a central symmetrical hexagon with obtuse angles or a rectangle,

  2. 2.

    the area of HH is equal to the length of the perpendicular projection of CC on the zz-axis,

  3. 3.

    p2+q2+r2=1p^{2}+q^{2}+r^{2}=1, where p,qp,q and rr are the third coordinates of the three non-parallel normal vectors of the faces of CC. Similar equations hold for the second and for the first coordinates.

Proof of Lemma 1.

Statement 1 follows from the visual geometric observation that in an infinite wedge bounded by two half planes with an acute angle, no solid cube can reach the edge of the wedge on its convex side. Indeed, if this would be possible, then a continuity argument would imply that the contact is possible also with one face of the cube on one of the bounding half planes, a contradiction.

Refer to caption
Figure 2: Basic properties of the projections of a cube.

Statement 2 is a well known property of the cube. Figure 2 becomes a proof without words, once it is noticed that the dot product of a unit edge vector and the vertical unit vector has two geometric meanings. On the one hand, it is equal to the area of the shadow of that face which is perpendicular to the unit edge vector. On the other hand, it is equal to the length of the perpendicular projection of the unit edge vector on the zz-axis.

Statement 3 must be also an old known property of the cube. Here we present a short proof via matrices. Let P,Q\overrightarrow{P},\overrightarrow{Q} and R\overrightarrow{R} be the three normal vectors with third coordinates p, q and r. Let AA be the matrix whose row vectors are these normal vectors. The linear transformation AA is an isometry so for all column vectors α,β3\alpha,\beta\in\mathcal{R}^{3}, (Aα)T(Aβ)=αTβ(A\alpha)^{T}\cdot(A\beta)=\alpha^{T}\cdot\beta. Equivalently, (αTAT)(Aβ)=αTβ(\alpha^{T}A^{T})\cdot(A\beta)=\alpha^{T}\cdot\beta. Thus, ATA=IA^{T}A=I, which in view of the row-column multiplication means that the column vectors of AA form an orthonormal basis. In particular, the third column vector is a unit vector, thus p2+q2+r2=1p^{2}+q^{2}+r^{2}=1. ∎

Next, we will show Lemma 2, which not only claims that the interior of every hexagonal shadow of a unit cube contains a unit square, but also explains, where such squares are located within a shadow.

Lemma 2.

Let HH be a hexagonal projection of a unit cube (Figure 3). Then, there are at least two pairs of opposite vertices of the hexagon HH so that each of these four (or six) vertices can share a vertex of a unit square in BB. Moreover, each of these four (or six) squares are unique and have the following properties:

  • The two vertices adjacent to the ’corner’ vertex, lie on a pair of opposite sides of HH.

  • The fourth vertices of these squares belong to interior of HH.

  • Each of these squares can be moved in the interior of the hexagon HH.

Remark.

First we used the GeoGebra software to see how does the shadow of a cube change when a cube is rotated in 33-space. The experiment made us believe that there might exists a unit square in every shadow, so that the sqaure shares a vertex with the projection of the cube. The conjectured special position allowed us to get the affirmative answer by carrying out a straightforward four page long computation, using two variables only. The simple proof bellow is different, it is adjusted to the box version and it reveals more geometric reasons.

Refer to caption
Figure 3: Every projection of a unit cube contains a unit square.
Proof of Lemma 2.

Let AA be one of the vertices of the hexagonal projection HH of the cube. Let VV be that vertex of the cube whose vertical projection is AA. Let us label the three unit edge vectors emanating from VV by P,Q\overrightarrow{P},\overrightarrow{Q} and R\overrightarrow{R}. If p,qp,q and rr denote the absolute values of the zz-coordinates of these vectors, then the lengths of the projections of these vectors are 1p2,1q2\sqrt{1-p^{2}},\sqrt{1-q^{2}} and 1r2\sqrt{1-r^{2}}. Let us orient the hexagon HH so that one pair of opposite sides is horizontal and the lengths of these sides equal to 1p2\sqrt{1-p^{2}}. Since all angles of HH are obtuse (Part 1 of Lemma 1) we may assume the labelling of side lengths and the areas of the sub-parallelograms are as shown on Figure 3.

In order to do a little computation, we introduce an xyxy-coordinate system centered at AA so that the horizontal line is the xx coordinate axis. Let BB^{\prime} and CC^{\prime} be the adjacent vertices of AA in HH. Choose points BB and CC on the horizontal sides of HH at distance 11 from AA. Let BB the point at distance bb from BB^{\prime} and BB^{\prime} at distance bb^{\prime} from the yy-axis. Distances cc and cc^{\prime} are introduced analogously for points CC and CC^{\prime}. We have

(b+b)2=12(r1p2)2=1r21p2=q21p2(b+b^{\prime})^{2}=1^{2}-(\frac{r}{\sqrt{1-p^{2}}})^{2}=1-\frac{r^{2}}{1-p^{2}}=\frac{q^{2}}{1-p^{2}},

(b)2=1q2(r1p2)2=q21p2q2=p2q21p2(b^{\prime})^{2}=1-q^{2}-(\frac{r}{\sqrt{1-p^{2}}})^{2}=\frac{q^{2}}{1-p^{2}}-q^{2}=\frac{p^{2}q^{2}}{1-p^{2}},

b=q1p2(1p)=q1p1+pb=\frac{q}{\sqrt{1-p^{2}}}(1-p)=q\sqrt{\frac{1-p}{1+p}}.

Similar expression holds for cc: c=r1p1+pc=r\sqrt{\frac{1-p}{1+p}}.

Thus, we have the following three equivalent inequalities,

b+c<(1p2)(q+r)1p1+p<1p2q+r<1+pb+c<\sqrt{(1-p^{2})}\Leftrightarrow(q+r)\sqrt{\frac{1-p}{1+p}}<\sqrt{1-p^{2}}\Leftrightarrow q+r<1+p.

Let DD be the vertex which completes A,B,CA,B,C to a square. Since both HH and ABCDABCD are centrally symmetric, DD lies on the horizontal line through the vertex of HH opposite to AA. Now, the first and the last equivalent inequalities mean that DD is inside of HH if q+r1+pq+r\leq 1+p. Since all p,q,r(0,1)p,q,r\in(0,1), this holds, if pp is the largest or the second largest among p,q,rp,q,r, which proves the first two properties listed in Lemma 2. A small shift followed by a small rotation of these unit square ensures that there is a unit square which lies in the interior of the hexagon. ∎

Definition.

An infinite vertical cylinder, which is raised over a horizontal rectangle will be called rectangular tube.

We will need

Lemma 3.

Every planar cross section of a rectangular tube contains a rectangle congruent to its base.

Remark.

This lemma is a special case of a very strong theorem of Kós and Törőcsik [8]. They proved that every convex disc covers its shadow. Here we present a new short proof for Lemma 3, so that this paper remains self contained.

Refer to caption
Figure 4: The plane of cross section folded over the horizontal plane.
Proof of Lemma 3.

Let RR be a rectangle of sides aa and bb, aba\leq b. If the plane of the cross section is horizontal, then Lemma 3 is obviously true. Otherwise, denote the angle between the plane of the tube’s base SS and the plane of the cross section PP by α\alpha (Figure 4). PP is a parallelogram with opposite sides at least at distances aa and bb respectively. Introduce an x,yx,y coordinate system in the horizontal plane of RR so that the xx axis belongs also to the plane of PP. Let us fold the plane of the cross section into the horizontal plane around the xx-axis. It is clear that there is a congruent image PP^{*} of PP so that the linear map (x,y)(x,1cosαy)(x,y)\rightarrow(x,\frac{1}{\cos\alpha}y) takes the rectangle RR to parallelogram PP^{*}. We denote the image of a point pp under this mapping by pp^{*}.

Let us label the vertices of rectangle RR by 1,2,31,2,3 and 44 in the order of their yy coordinates (Figure 1). It is easy to see that 1cosα>1\frac{1}{\cos\alpha}>1 implies that PP^{*} has obtuse angles at vertices 22^{*} and 33^{*}. Depending on how the diagonal 232^{*}3^{*} splits the obtuse angle at 22^{*}, we distinguish two cases:

Case 1. Both sub-angles at 22^{*} are greater than equal to the corresponding sub-angles at 22. In this case a centrally positioned copy of RR inside of PP^{*}, so that one of the diagonal of RR lies on 232^{*}3^{*}, is contained in the parallelogram PP^{*}.

Case 2. One of the sub-angles, say 324\angle 3^{*}2^{*}4^{*}, is less than the corresponding sub-angle at 22. In this case, the length of the side 242^{*}4^{*} is greater than the length of side 2424, which is aa. Since the distance between the sides 242^{*}4^{*} and 131^{*}3^{*} is at least bb, the rectangle, which has a vertex at 22^{*} and has a side of length aa on the segment 242^{*}4^{*} is contained in the parallelogram PP^{*}. ∎

4 Proof of Theorems

Proof of Theorem 1.

Let BB be a rectangular box with edge lengths abca\leq b\leq c. Shorten the longest edges of BB to bb. It is enough to show that the new box BB^{\prime} contains a rectangle of dimensions a,ba,b. Let hexagon HH be a shadow of BB^{\prime}. Two of the six sides of HH are projections of edges of length aa, while the remaining four sides are projections of edges of length bb. To depict this property we label the sides and their lengthes with pr(a)pr(a) or pr(b)pr(b) on Figure 5. We have two pairs of opposite vertices which are endpoints of pr(a)pr(a) sides. In view of Lemma 2 at least one of these two pairs have vertices, say AA and AA^{\prime}, so that moving a cube with edge length aa to the corner of the pre-image of AA, the projection of the cube contain a square whose vertex is AA. Now it is easy to see that one can extend one pair of parallel sides of this square to bb so that the new rectangle is still contained in HH, what we wanted. A small shift followed by a small rotation of this rectangle ensures that there is a unit square which lies in the interior of the hexagon. ∎

Refer to caption
Figure 5: Every projection of a box contains a copy of the smallest face.
Proof of Theorem 2.

Assume that a homothetic box λB\lambda B can be passed through a stationary box BB with sides abca\leq b\leq c. Although Theorem 2 is obviously true if λ1\lambda\leq 1, we present an argument which works for all λ\lambda. On Figure 6, which illustrates the proof, we choose a λ\lambda much smaller than 11 so that the figure becomes less crowded and the proof can be better understood. Let H1H_{1} be the shadow of the stationary rectangular box BB and HλH_{\lambda} be the shadow of the traveling homothetic rectangular box λB\lambda B. We proved already that the hexagon HλH_{\lambda} contains a rectangle λR\lambda R of sides λa\lambda a and λb\lambda b. Since both H1H_{1} and λR\lambda R are central symmetric, the rectangle λR\lambda R can be shifted in center position while remaining inside of the hexagon H1H_{1}. Now, choose two adjacent vertices, say UU and VV, of this relocated rectangle. Since U,VU,V are inside of the shadow of BB, we can choose two points UU^{*}, VV^{*} inside of the box BB directly over UU and VV. Finally, reflect UU^{*}, VV^{*} through the center of the box BB two get a parallelogram PP in box BB. PP’s shadow is the rectangle of sides aa and bb. According to Lemma 3, parallelogram PP contains a rectangle of sides aa and bb. ∎

Refer to caption
Figure 6: Proof of Theorem 2.

References

  • 1. Y. Chai, L. Yuan and T. Zamfirescu, Rupert properties of archimedean solids, Amer. Math. Monthly 125:6 (2018), 497–504.
  • 2. M. Gardner, The colossal book of mathematics, W.W. Norton, New York, 2001.
  • 3. A. Ehrenfeucht, The cube made interesting, Pergamon Press, Oxford, 1964.
  • 4. G. Huber, K.P. Schultz and J.E. Wetzel, The n-cube is Rupert Amer. Math. Monthly 125:6 (2018), 505–512.
  • 5. R.P. Jerrard and J.E. Wetzel, Universal stoppers are Rupert College Math. J. 39:2 (2008), 90–94.
  • 6. R.P. Jerrard and J.E. Wetzel, Prince Rupert’s rectangles Amer. Math. Monthly 111 (2004), 22–31.
  • 7. R.P. Jerrard, J.E. Wetzel and L. Yuan, Platonic passages Math. Mag. 90:2(2017), 87–98.
  • 8. G. Kós and J. Törőcsik, Convex discs can cover their shadows Discr. Comp. Geom. 5 (1990), 529–531.
  • 9. V.F. Rickey, Dürer’s Magic Square, Cardano’s Rings, Prince Rupert’s Cube, and Other Neat Things MAA Short Course ”Recreational Mathematics”, Albuquerque, New Mexico, 23 August (2005).
  • 10. D.J.E. Schrek, Prince Rupert’s problem and its extensions by Pieter Nieuwland Scripta Math. 16 (1950), 73–80, 261–267.
  • 11. C.J. Scriba, Das problem des Prinzen Ruprecht von der Pfalz Praxis der Math. 10(9) (1968), 241–246.
  • 12. J.H. Van Swinden, Grondbeginsels der Meetkunde, 2nd ed., Amsterdam, 1816, 512–513.
  • 13. J.E. Wetzel, Rectangles in rectangles Math. Mag. 73 (2000), 204–211.