Introduction

Partitioning a space with points into polyhedra in such a way that each polyhedron encloses exactly one point and then characterizing the polyhedral tiling is a promising strategy to study a wide range of structures1,2,3,4,5,6,7,8,9,10,11,12. For example, in studying the atomic structure of a material, the space can be divided into the so-called Voronoi polyhedra1,2,3,4,5,6,7,8,9, where each polyhedron encloses its associated atom. By using this method, for example, a way of how an atom X is surrounded by its first and second nearest-neighbour atoms is represented by the local tiling structure composed of the Voronoi polyhedra associated with the atom X and its first nearest-neighbour atoms.

Since such a local tiling structure can be regarded as a part of a four-dimensional polytope (4-polytope) called a polychoron, a method to describe how a polychoron is constructed from its building-block polyhedra can be used to study the structure of materials. For this reason, we have recently developed a theory of polytopes13 that is based on the hierarchy of structures of polytopes14,15,16,17,18: a polyhedron (3-polytope) is a tiling by polygons (2-polytopes), a polychoron (4-polytope) is a tiling by polyhedra (3-polytopes), and so on. Specifically, we have first created the p3 -code for representing polyhedra. The p3 -code consists of (1) an encoding algorithm for converting a way of how polygons are arranged to form a polyhedron into a p3 -codeword (p3 for short) and (2) a decoding algorithm for recovering the original polyhedron from its p3. By generalizing the p3 -code, we have created the p4 -code for representing polychora. By using the p4 -code, a way of how polyhedra are arranged to form a polychoron can be converted into a p4 -codeword (p4 for short), from which the original polychoron can be recovered. A polyhedral tiling can be characterized by distribution of p4s of local tiling structures of different central polyhedra. However, p4 is redundant as described below.

The p4-codeword contains p3(1), p3(2), p3(3),  , and p3(C), where p3(i) is p3 of the polyhedron i and C is the number of polyhedra of the polychoron. Each p3(i) contains p2(i1), p2(i2), p2(i3), , and p2(iF(i)), where each p2(ij) is the number of edges of the face j of the polyhedron i and F(i) is the number of faces of the polyhedron i. Here, we point out that p2(ij)s of all the faces of all the polyhedra are recorded in p4. However, since polyhedra are glued together face to face, the pair of faces glued each other have the same number of edges. p4 is thus redundant and lengthy. For example, if the face y of the polyhedron x is glued to the face w of the polyhedron v, then p2(vw) = p2(xy), so that p2(vw) in p4 is redundant.

Redundant codewords mean the lack of knowledge of structures of polychora. In addition, redundant codewords are practically unfavourable for both humans and computers. For humans, recognizing and writing down lengthy codewords are troublesome. For computers, larger hard drives are necessary to store codewords and more computation time is necessary to determine the equivalence of codewords.

In this paper, we develop a theory to reduce the redundancy in p4. For this purpose, we exploit the fact that the polyhedra are glued together face to face. Specifically, we regard two-dimensional regions shared by faces of adjacent polyhedra as building blocks of a polychoron. To distinguish between parts of a polychoron and parts of a polyhedra, we refer the two-dimensional building blocks of a polychoron to ridges. As the distinction between edges of a polyhedron and sides of a polygon was crucial for L. Euler to find his famous polyhedral formula, V − E + F = 214, the distinction between ridges of a polychoron and faces of a polyhedron is crucial for our theory. To represent a polychoron using ridges, we formulate a method to convert p4 into , where the superscript “rs” indicates the ridge-sequence. Note that p4 instructs how to construct a polychoron from its building-block polyhedra, while instructs how to construct a polychoron from its building-block ridges. The length of is as short as half of p4. By generalizing the method to higher-dimensional polytopes, we develop a unified theory of how a polytope is constructed from its two-dimensional components.

Results

Bare essentials of the p4-code

We will formulate the -code consisting of (1) an encoding algorithm for converting p4 into and (2) an decoding algorithm for recovering the original polychoron or p4 from . We start with the brief explanation of bare essentials of the p4-code needed to formulate the -code. Specifically, we explain how to recover a polychoron from p4. For reader’s convenience, the encoding algorithm is described in Supplementary Note. The full details of the p4-code has been given in the previous paper13. Since polychora associated with disordered structures are simple, we deal with simple polychora. By a simple polychoron, we mean a polychoron whose 0-faces are all incident with four peaks. Here, 0-faces and peaks are zero- and one-dimensional components of a polychoron, respectively. Since a simple polychoron is composed of simple polyhedra, we first explain the p3-code for simple polyhedra. By a simple polyhedron, we mean a polyhedron whose vertices are all incident with three edges.

A polyhedron can be regarded as a tiling by polygons of the surface of a three-dimensional object that is topologically the same as a sphere. According to the idea developed by L. Euler, A. M. Legendre, F. Möbius, and P. R. Cromwell14, we assume that polygons are glued such that (1) any pair of polygons meet only at their sides or corners and that (2) each side of each polygon meets exactly one other polygon along an edge. We stress that we distinguish between parts of a polyhedron and those of the building-block polygons. Specifically, vertices and edges are zero- and one-dimensional parts of a polyhedron, respectively. On the other hands, corners and sides are zero- and one-dimensional parts of a polygon, respectively. Since this idea plays a central role in our theory, we need a verb to briefly describe the relation between parts of a polyhedron and those of polygons. For this purpose, we use the verb “contribute”. For example, when we say that the corners contribute to the vertex or the vertex is contributed by the corners, we mean that the vertex is a point on a polyhedron at which the corners of polygons meet. We also say that a polygon (side) contributes to a vertex if one of its corners (endpoints) contributes to the vertex. Similarly, when we say that the edge is contributed by the sides, we mean that the edge is a line segment on a polyhedron along which the sides of polygons meet. The face of a polyhedron is a polygon. But when we call a polygon, we regard it as a building block of a polyhedron. So, we may say the edge of a face. But we cannot say the edge of a polygon.

Using the p3-code, a way of how polygons are arranged to form a polyhedron can be converted into p3, which instructs how to construct the polyhedron from its building-block polygons. The p3-codeword consists of the polygon-sequence codeword (ps2) and the side-pairing codeword (sp), and is denoted as

where “;” is a separator. The ps2-codeword is denoted as

Here, p2(i) is the number of sides of the polygon i, and F is the number of polygons of the polyhedron. We note that the number of sides of the polygon i is identical with the number of edges of the face i.

If we know all information of p2(i)s and all information about which side should be glued to which side, we can construct a polyhedron by gluing polygons side to side. The ps2-codeword contains not only all information of p2(i)s, but also all or almost all information about which side should be glued to which side. Many polyhedra are represented just by ps2, but there are some polyhedra that need additional information about which side should be glued to which side. Such additional information is recorded in sp, which is denoted as

Here, y(i) and x(i) are the identification numbers (IDs) of sides. The pair of sides y(i) and x(i) is what we call a non-curable additional pair (side-na-pair y(i)x(i)). By a side-na-pair y(i)x(i), we mean that the sides y(i) and x(i) should be glued together. Here, y(i) > x(i) and y(i) < y(i + 1). Ns is the number of side-na-pairs.

Decoding p3 is constructing its original polyhedron by gluing together polygons side to side. To instruct which side should be glued to which side, we assign IDs to sides. We assign ij to the jth side of the polygon i, and the side-ID ij represents an integer: . In constructing a polyhedron, if a side of a polygon of the partial polyhedron is not glued to the other polygon, we call it a dangling side. We abbreviate the smallest-ID dangling side as the s-side. We regard that an isolated corner as well as two corners meeting at a point forms a vertex of a partial polyhedron. We also regard that a dangling side forms an edge. If the pair of dangling sides contribute to a vertex that is also contributed by three polygons, that vertex is said to be illegal. When an illegal vertex (i-vertex) is generated, we rectify it by gluing together the two dangling sides contributing to it. The polyhedron can be recovered from ps2;sp as follows:

1. (a) The polygon 1 is a p2(1)-gon.

(b) Assign IDs to its sides in a clockwise (CW) direction.

2. (a) The next polygon i (2 ≤ i ≤ F) is a p2(i)-gon.

(b) Assign IDs to its sides in a CW direction.

(c) Glue the side i1 to the s-side of the partial polyhedron.

(d) If y(n) (1 ≤ n ≤ Ns) is the side ID of the polygon i, then glue the side y(n) to the side x(n) of the partial polyhedron.

(e) If i-vertices are generated, then rectify them, and repeat this procedure until no i-vertices remain.

3. (a) Repeat the procedure 2 until all polygons are placed.

The edge IDs are assigned as follows. Given that each edge is contributed by two sides, we tentatively assign the smaller side ID to the edge, and then relabel IDs so that the edge i is the one with the ith smallest tentative ID.

We note that the p3-code can be used to represent a tiling by polygons of a torus without modification. But to represent a toroidal polyhedron, we need to specify how to embed the torus in the 3-dimensional Euclidean space to form a toroidal polyhedron. The p3-code can also be generalized to non-orientable planes such as the Klein bottle19 by defining the clockwise direction for the polygon i, in which IDs are assigned to sides, depending on the clockwise direction for the polygon to which the side i1 is glued.

The p3-code is generalized to the p4-code for polychora as follows. We regard a polychoron as a tiling by polyhedra of the surface of a four-dimensional object that is topologically the same as a 3-sphere. We assume that polyhedra are glued together such that (1) any pair of polyhedra meet only at their faces, edges, or vertices and that (2) each face of each polyhedron meets exactly one other polyhedron along a ridge. We distinguish parts of a polychoron and parts of its building-block polyhedra. The 0-face, peak, and ridge are a point, line segment, and area on a polychoron, where the vertices, edges, and faces of polyhedra meet, respectively. The cell of a polychoron is a polyhedron.

The p4-codeword consists of a polyhedron-sequence codeword (ps3) and a face-pairing codeword (fp), and is denoted as

Here,

C is the number of polyhedra of the polychoron. p3(i) = ps2(i);sp(i) is p3 of the polyhedron i.

The fp-codeword consists of what we call face-na-pairs wzv, and is denoted as

Here, w(i) and v(i) are face IDs. w(i) > v(i) and w(i) < w(i + 1). z(i) is the global side ID of a side of the polygon w(i). The global side IDs will be explained latter. By a face-na-pair w(i)z(i)v(i), we mean that the faces w(i) and v(i) should be glued together in such a way that the edge of the face w(i) contributed by the side z(i) is glued to the smallest-ID edge of the face v(i). Nf is the number of face-na-pairs of the polychoron. Note that, in order to formulate , fp of the present work is slightly modified from the original definition13. For the original definition, z(i) is the edge ID of the edge of the face w(i) glued to the smallest-ID edge of the face v(i).

In decoding p4, if a face of a polyhedron of the partial polychoron is not glued to the other face, we call it a dangling face. By the edge ij (face ij), we mean the jth edge (face) of the polyhedron i. We abbreviate the smallest-ID dangling face as the s-face. We regard that an isolated edge as well as two edges meeting along a line segment forms a peak of a partial polychoron. We also regard that a dangling face forms a ridge. In a partial polychoron, if the pair of dangling faces contribute to a peak that is also contributed by three polyhedra, we call that peak an illegal peak (i-peak). When an i-peak is generated, we rectify it by gluing together the two dangling faces contributing to it. The polychoron can be recovered from ps3;fp as follows:

1. (a)  Decode p3(1) to obtain the polyhedron 1, assigning face and edge IDs.

2. (a) Decode p3(i) to obtain the next polyhedron i (2 ≤ i ≤ C), assigning face and edge IDs.

(b) Glue the face i1 to the s-face of the partial polychoron in such a way that the edge i1 is glued to the smallest-ID edge of the s-face.

(c) If w(n) (1 ≤ n ≤ Nf) is the face ID of the polyhedron i, then glue the face w(n) to the face v(n) of the partial polyhedron in such a way that the edge of the face w(n) contributed by the side z(n) is glued to the smallest-ID edge of the face v(n).

(d) If i-peaks are generated, then rectify them, and repeat this procedure until no i-peaks remain.

3. (a) Repeat the procedure 2 until all polyhedra are placed.

Ridge IDs are assigned as follows. Given that each ridge is contributed by two faces, we tentatively assign the smaller face ID to the ridge. We call the ID thus assigned the tentative ridge ID. We then relabel IDs so that the ridge i is the one with the ith smallest tentative ID. The tentative ridge IDs and ridge IDs play a key role in reducing the redundancy in p4. Peak IDs are also assigned similarly.

Preliminary arrangements

An outline of converting p4 into is illustrated in Fig. 1. We will first break p4 down into its ps2s and sps, and reconstruct , which provides us a good perspective for reducing the redundancy. We will then reduce the redundancy by converting into . Finally, to make our theory more beautiful, we unify spg* and fp into a part-pairing codeword (pp), and obtain .

Figure 1: Overview of the -code.
figure 1

Three-dimensional Schlegel diagrams13,15,18 (a projection from four- to three-dimensional space) are used to illustrate the polychoron. Note that the interior of the polyhedron abcd on the polychoron in four-dimensional space (not shown) is mapped to the exterior of the outside polyhedron abcd on the Schlegel diagram.

Full size image

To formulate , we distinguish local IDs and global IDs. When we call the polygon j of the polyhedron i, the number j is what we call the local polygon ID associated with the polyhedron i. We can designate the same polygon as the polygon ij. The symbol ij is what we call the global polygon ID of the polychoron. The symbol ij also represents the number: , where F(k) is the number of polygons of the polyhedron k. Similarly, by the side mn of the polyhedron i, we mean the nth side of the polygon m of the polyhedron i. The number n is the local side ID associated with the polygon m of the polyhedron i, while the symbol mn is the local side ID associated with the polyhedron i and is also the number, . Here, p2(ik) is the number of sides of the polygon ik. Using the global side ID, we can designate the side mn of the polyhedron i as the side . The symbol also represents the number: , where E(k) is the number of edges of the polyhedron k.

The sp(i)-codeword of p3(i) = ps2(i);sp(i) is written using the local side IDs associated with the polyhedron i. Using the global side IDs, we rewrite p3(i) as . Here, is obtained by translating from local side ID into global side ID. . y(ij) and x(ij) are the local side IDs for the jth side-na-pair of the polyhedron i. Ns(i) is the number of side-na-pairs of the polyhedron i. For example, p3(i) = 46565475543; 7444 = 46565475543;34 19 is rewritten as . Reversely, p3(i) can be recovered from just by translating the global side IDs iyix into their corresponding local side IDs yx.

By putting together spg(i)s, spg* is defined as follows:

Here,

Similarly, is defined by putting together ps2(i)s as follows:

In the last transformation, we translated the symbols ij into the corresponding numbers. R is the number of ridges of the polychoron.

By putting , spg*, and fp together, . For example, for a polychoron A shown in Fig. 1,

We can recover p4 from as follows. By construction, the first F(1) digits of form ps2(1). However, we do not know F(1) beforehand. To find out F(1), we regard as p3, and decode it until a polyhedron is completed. Suppose that a polyhedron is completed when the αth digit of is decoded. Then F(1) = α, and . If the sides of the polyhedron are found in spg*, record them in spg(1). We then remove ps2(1) from , and obtain . As with , the first F(2) digits of form ps2(2). To find out F(2), we decode using the p3-code. Suppose that a polyhedron is completed when the βth digit of is decoded. Then F(2) = β, and . If the sides of the polyhedron are found in spg*, record them in spg(2). We then remove ps2(2) from , and obtain . By repeating this procedure, we can determine , and therefore p4. As an example, we illustrate the procedures of recovering p4[A] from in Supplementary Note.

Reveal and remove redundancy in

To reveal the redundancy in , we observe how the polychoron A shown in Fig. 1 is recovered from . After determining p4[A], we construct the polyhedron 1 (3333) and polyhedron 2 (34443), and then glue them together in such a way that the face 21 is glued to the face 11. Therefore, p2(21) must be equal to p2(11). Thus, p2(21) is redundant. Next we attach the polyhedron 3 (34443) to the partial polychoron in such a way that the face 31 is glued to the face 12 of the partial polychoron. Therefore, p2(31) must be equal to p2(12), and p2(31) is redundant. In addition, to rectify an i-peak, we glue together the faces 32 and 22. Therefore, p2(32) must be equal to p2(22), and p2(32) is redundant.

In general, when two faces ij and mn (m > i) are glued together, p2(mn) is redundant, while p2(ij) is essential. Since the face mn meets the face ij along the ridge with the tentative ID ij, p2(mn) = p2(ij) = rt(ij). Here, rt(xy) is the number of peaks of the ridge with the tentative ID xy. Thus, the number of peaks of every ridge is doubly recorded in .

Returning to the polyhedron A, we will remove all the redundant p2(mn)s from and construct the sequence of essential p2(ij)s,

The sequence of essential p2(ij)s is identical with the sequence of rt(ij)s,

By rewriting the sequence using the ridge IDs (not tentative ridge IDs), we obtain what we call the ridge-sequence codeword (rs),

Here, r(i) is the number of peaks of the ridge with the ID i. The number of peaks of every ridge is recorded just once in rs, and the redundancy is thus reduced. In general, the redundancy can be reduced by modifying p4 into Here, rs = r(1)r(2)r(3)  r(R).

How to recover p4 from

The -codeword contains information about how to assemble ridges to form a polychoron in the sense that p4 can be recovered from . As is summarized in Fig. 2, to recover p4, we determine step-by-step. To determine p3(i), we deduce step-by-step. To deduce p2(ij), we examine whether the face ij should be glued to an existing face of the partial polychoron or create a new ridge.

Figure 2
figure 2

Procedures for recovering p4 from .

Full size image

We first describe how to determine p3(1) = ps2(1);sp(1) from . All faces of polyhedron 1 create new ridges. By construction, the first F(1) digits of r(1)r(2)r(3)  r(R) form ps2(1). However, we do not know F(1) beforehand. To find out F(1), we regard r(1)r(2)r(3)  r(R);spg* as p3, and decode it until a polyhedron is completed. The polyhedron thus obtained is the polyhedron 1. Suppose that a polyhedron is completed when the αth face is decoded. Then F(1) = α, and ps2(1) = r(1)r(2)r(3)  r(α). Every time we recover a polygon in the decoding process, we search spg* for side-na-pairs of the polyhedron 1. If side-na-pairs are found, we record their corresponding local side IDs in sp(1). By combining ps2(1) and sp(1), we obtain p3(1) = ps2(1);sp(1).

For 2 ≤ i, p3(i) can be determined from and p3(1)p3(2)p3(3)  p3(i − 1). Our first task is to deduce p2(i1). For this purpose, we construct a partial polychoron D4(i − 1), which is obtained by decoding p3(1)p3(2)p3(3)  p3(i − 1);fp. Since the face 1 of the polyhedron i should be glued to the s-face of D4(i − 1), p2(i1) = p2(IDs−face(i − 1)). Here, IDs−face(k) is the global face ID of the s-face of D4(k).

For 2 ≤ j, p2(ij) can be determined from , p3(1)p3(2)p3(3)  p3(i − 1), and p2(i1)p2(i2)p2(i3)  p2(ij−1). To deduce p2(ij), we examine whether the face ij should be glued to an existing face or create a new ridge. We first search the w part of fp for ij. If ij is found, the faces ij( = w) and v form a face-na-pair. Since those faces should be glued together, p2(ij) = p2(v). If ij is not found, we construct a partial polyhedron D3(ij−1), which is obtained by decoding p2(i1)p2(i2)p2(i3)  p2(ij−1); using the p3-code. We then glue the face i1 (of D3(ij−1)) to the s-face of D4(i − 1) in such a way that the edge i1 is glued to the smallest-ID edge of the s-face. If w(n) (1 ≤ n ≤ Nf) is the face ID of D3(ij−1), then glue the face w(n) to the face v(n) of D4(i − 1) in such a way that the edge of the face w(n) contributed by the side z(n) is glued to the smallest-ID edge of the face v(n). If i-peaks are generated, then we rectify them until no i-peaks remain. We write D3(ij−1) & D4(i − 1) for the partial polychoron thus obtained. The peak contributed by the s-side of D3(ij−1) plays a key role in determining whether the face ij should be glued to an existing face or create a new ridge, therefore we call that peak the key peak (k-peak). We write IDk−peak(ij−1) for the global peak ID of the k-peak. There exists one dangling face contributing to the k-peak, which we call a candidate face (c-face). We write IDc−face(ij−1) for the global face ID of the c-face. The face ij will be glued to the face IDc−face(ij−1) or create a new ridge. Now, we need to consider two cases:

(case 1)  The k-peak of D3(ij−1) & D4(i − 1) is contributed by three polyhedra (D3(ij−1) and two from D4(i − 1)). In this case, in constructing D3(ij) & D4(i − 1), when the peak IDk−peak(ij−1) is contributed by three polyhedra and the face ij is not glued to the face IDc−face(ij−1), that peak will be illegal. To rectify the i-peak, the faces ij and IDc−face(ij−1) should be glued together. Thus, p2(ij) = p2(IDc−face(ij−1)).

(case 2)  The k-peak of D3(ij−1) & D4(i − 1) is contributed by two polyhedra (D3(ij−1) and one from D4(i − 1)). In this case, the face ij should create a new ridge. Thus, p2(ij) = r(Nridge(ij−1) + 1). Here, Nridge(mn) is the number of ridges of D3(mn) & D4(m − 1).

Part-pairing codeword and

To make our theory more beautiful, we modify the way to record na-pairs. The side-na-pairs yx are recoded in spg*, while the face-na-pairs wzv are in fp. To distinguish them, there is a separator “;” between spg* and fp as . To remove the separator, we will unify spg* and fp into pp. Finally, we will obtain . As a result, both polyhedra and polychora are represented by codewords of the same format, namely two number sequences separated by “;”.

To formulate pp, we introduce the notion of parts. We regard the sides of a polygon are parts of that polygon. We also regard the polygon itself is the part of that polygon. We define the set of parts of the polygon i as

Similarly, for j > 2, we define the set of parts of the polyhedron j as

We assign IDs to parts such that we can identify side-na-pairs yx and face-na-pairs wzv in recovering the original polychoron. To meet this requirement, we assign IDs to parts of the polychoron in the order of S[polygon 1], polyhedron 1, S[polygon 2], …, S[polygon F(1)], edges of polyhedron 1, polychoron 1, S[polyhedron 2], …, S[polyhedron C], ridges of polychoron 1, peaks of polychoron 1.

The pp-codeword is obtained as follows. We first translate spg*;fp into part ID. Then we remove the separator “;”. Finally, we obtain

The side- and face-na-pairs can be identified from pp as follows. Let p(i) be the ith digit of pp. If the part p(i) is a side Y, the part p(i + 1) is a side X, and Y > X, then the pair p(i)p(i + 1) is a side-na-pair. If the part p(i) is a face W and the part p(i + 2) is a face V, then the pair p(i)p(i + 1)p(i + 2) is a face-na-pair.

Note that the amount of tasks needed to generate is comparable to that needed to generate p4. This is because converting p4 to amounts to just assigning IDs to ridges and parts of the polychoron. We also note that the length of is shorter than that of p4 by R*, where R* is the number of ridges contributed by two faces. R* = R for a polychoron, while R* < R for a partial polychoron. Therefore, the compression efficiency gets worse for partial polychora. As described above, by converting 3333 34443 34443 34443 34443 3333 into 33334443443433, data of the polychoron is compressed to half. On the other hand, for example, p4 and of a partial polychoron composed of one 3333-polyhedron and one 34443-polyhedron are 3333 34443 and 33334443, respectively. Just one number “3” is removed by converting p4 into .

The p4-codeword can be recovered from by modifying the procedure for recovering p4 from as follows (Fig. 3):

  1. 1

    Determine p3(1) = ps2(1);sp(1) as follows:

    1. a

      Decode r(1)r(2)r(3) ··· r(R);pp using the the p3-code.

    2. b

      If a polyhedron is completed when the αth face is decoded, then ps2(1) = r(1)r(2)r(3) … r(α).

    3. c

      If side-na-pairs of the polyhedron are found in pp, record their corresponding local side IDs in sp(1).

  2. 2

    Determine the next p3 = ps2(i);sp(i)(2 ≤ i) as follows:

    1. a

      p2(i1) = p2(IDs-face(i − 1)).

    2. b

      To determine the next p2(ij) (2 ≤ j), we search pp for the part ID of the face ij. Here, two cases arise:

      1. I

        If the part p(k) is the face ij and the part p(k + 2) is a face, then let mn be the face-ID of the part p(k + 2), and p2(ij) = p2(mn).

      2. II

        Otherwise, we examine the k-peak, and then additional two cases arise:

        1. i

          If the k-peak is contributed by three polyhedra, then p2(ij) = p2(IDc-face(ij−1)).

        2. ii

          Otherwise, p2(ij) = r(Nridge(ij−1) + 1).

    3. c

      Decode p2(i1)p2(i2)p2(i3) … p2(ij);pp using the the p3-code. Two cases then arise:

      1. I

        If a polyhedron is completed, then ps2(i) = p2(i1)p2(i2)p2(i3) … p2(ij). If side-na-pairs of the polyhedron are found in pp, record their corresponding local side IDs in sp(i). p3(i) is thus determined.

      2. II

        Otherwise, repeat the procedure 2b.

  3. 3

    Decode p3(1)p3(2)p3(3) … p3(i);pp using the the p4-code. Two cases then arise:

    1. a

      If a polychoron is completed, then ps3 = p3(1)p3(2)p3(3)  p3(i). If face-na-pairs are found in pp, record their corresponding global face, side, and face IDs in fp. Thus, p4 = ps3;fp is determined.

    2. b

      Otherwise, repeat the procedure 2.

Figure 3: Procedures for recovering p4 from .
figure 3

The differences from the algorithm for is highlighted in yellow.

Full size image

As an example, we illustrate how to recover p4[A] from as follows:

  1. 1

    We decode rs[A] using the p3-code. When the 4th digit is decoded, a 3333-polyhedron is obtained, thereby it turns out p3(1) = 3333 (Fig. 4).

    Figure 4: How to determine p3(1) from 33334443443433.
    figure 4

    The dashed lines are the edges contributed by one polygon. The solid lines are the edges contributed by two polygons. Each s-side to which the next polygon is glued is coloured red. For the completed polyhedron 1, global edge IDs are shown near their edges. The polyhedron 1 is D4(1), and its s-face is coloured yellow.

    Full size image
  2. 2

    We determine p3(2) as follows:

    1. a

      The 3333-polyhedron is the partial polychoron D4(1). The s-face of D4(1) is the face 11 (Fig. 4). Since the face 1 of the polyhedron 2 will be glued to the face 11, p2(21) = p2(11) = 3.

    2. b

      We construct the partial polyhedron D3(21), glue it to the partial polychoron D4(1), and obtain the partial polychoron D3(21) & D4(1) (Fig. 5). Since the k-peak ab is contributed by two polyhedra (polyhedron 1 and D3(21)), the face 22 will create a new ridge. Since has four ridges abc, bad, cbd, and acd, Nridge(21) = 4. Thus, p2(22) = r(Nridge(21) + 1) = r(5) = 4.

      Figure 5: Partial polychoron D3(21) & D4(1).
      figure 5

      D3(21), D4(1), and D3(21) & D4(1) are illustrated using three-dimensional Schlegel diagrams. The polyhedron abcd is the outside polyhedron. The dotted and dashed bold lines of the partial polychora indicate peaks contributed by one and two polyhedra, respectively. The face 21 (of D3(21)) and the s-face of D4(1) (face 11) are coloured blue. By gluing together the blue faces, D3(21) & D4(1) is obtained. The s-side b′a′ of D3(21) (red dashed line) contributes to the peak ab of D3(21) & D4(1) (red-bold-dashed line), so that the peak ab is the k-peak. The dangling face bad is the c-face, for it contributes to the k-peak. The c-face is coloured green. The k-peak is contributed by two polyhedra (polyhedron 1 and D3(21)). Global face and side IDs of D3(21) are shown near D3(21) for reference. Note that since the polyhedron 2 is an inside polyhedron, a counter CW direction such as b′ → a′ → c′ around the face b′a′c′ of the polyhedron 2 on the Schlegel diagram corresponds to a CW direction around the corresponding face of the polychoron in four-dimensional space.

      Full size image
    3. c

      For the same reason, p2(23) = r(6) = 4, p2(24) = r(7) = 4, p2(25) = r(8) = 3.

    4. d

      When we decode p2(21)p2(22)p2(23)p2(24)p2(25) = 34443, a polyhedron is completed, thereby it turns out p3(2) = 34443.

  3. 3

    We determine p3(3) as follows:

    1. a

      We construct D4(2) from the partial p4-codeword: p3(1)p3(2) = 3333 34443 (Fig. 6). The s-face of D4(2) is the face 12. Therefore, p2(31) = p2(12) = 3.

      Figure 6: Partial polychoron D4(2).
      figure 6

      By gluing the blue face 21 (of polyhedron 2) to the blue s-face of D4(1), D4(2) is obtained. The s-face of D4(2) (face 12) is coloured yellow.

      Full size image
    2. b

      We construct D3(31) & D4(2) by gluing D3(31) to D4(2) (Fig. 7). Since the k-peak is contributed by three polyhedra (polyhedra 1 and 2, and D3(32)), the face 32, should be glued to the face IDc−face(31) (face abfe). Thus, p2(32) = p2(IDc−face(31)) = p2(22) = 4.

      Figure 7: Partial polychoron D3(31) & D4(2).
      figure 7

      By gluing the blue face 31 (of D3(31)) to the blue s-face of D4(2), D4(31) & D4(2) is obtained. The red-bold-solid peak ab of D4(31) & D4(2) is the k-peak, for it is contributed by the s-side a′b′ of D3(31). The green dangling face 22 is the c-face, for it contributes to the k-peak. Three polyhedra (polyhedra 1 and 2 and D3(31)) contribute to the k-peak.

      Full size image
    3. c

      We construct D3(32) from the partial p3-codeword 34, glue it to D4(2), and obtain (Fig. 8). Since the k-peak is contributed by two polyhedra (polyhedron 1 and D3(32)), the face 33 will create a new ridge. Thus, p2(33) = r(Nridge(32) + 1) = r(9) = 4.

      Figure 8: Partial polychoron D3(32) & D4(2).
      figure 8

      The k-peak bd (red-bold-dashed line) is contributed by two polyhedra (polyhedron 1 and D3(32)).

      Full size image
    4. d

      For the same reason, p2(34) = r(10) = 4, and p2(35) = r(11) = 3.

    5. e

      When we decode p2(31)p2(32)p2(33)p2(34)p2(35) = 34443, a polyhedron is completed, thereby it turns out p3(3) = 34443.

  4. 4

    In a similar way, p3(4) is determined to be p2(13)p2(24)r(12)p2(33)r(13) = 34443. p3(5) = p2(14)p2(23)p2(34)p2(43)r(14) = 34443. p3(6) = p2(25)p2(35)p2(55)p2(45) = 3333.

  5. 5

    When we decode p3(1)p3(2)p3(3)p3(4)p3(5)p3(6), a polychoron is completed, thereby it turns out p4[A] = 3333 34443 34443 34443 34443 3333.

Generalization to higher dimensional polytopes

The p4-code can be generalized to the pn-code for n-polytopes (see Supplementary Note and Supplementary Table S1). The pn-codeword instructs how to construct the n-polytope from its building block (n − 1)-polytopes. However, as in the case of p4, pn is redundant. By reducing the redundancy, we can obtain . Here, is the n-dimensional generalization of . The superscript “fs2” indicates the 2-face-sequence codeword. The i-face is the i-dimensional face of an n-polytope. For example, a 2-face of a polychoron is a ridge, and a 1-face of a polychoron is a peak.

As an example, we explain pns for n-dimensional cubes (n-cubes), and then demonstrate how the pns are converted into their corresponding s. The 3-cube is an ordinary cube, and p3[3-cube] = 444444 = 46. The 4-cube is composed of eight 3-cubes, and p4[4-cube] = 46 46 46 46 46 46 46 46 = p3[3-cube]8. The 5-cube is composed of ten 4-cubes, and 10. In general, an n-cube consists of 2n (n − 1) − cubes20, and pn[n-cube] = pn−1[(n − 1) − cube]2n (for n ≥ 3).

The number of 1-faces of each 2-face of an n-polytope is (n − 2)! times recorded in pn, so that reducing the redundancy has a greater impact for higher dimensional polytopes. The redundancy can be reduced by using fs2 (see Supplementary Note and Supplementary Figure S1), which is denoted as

Here, f2(i) is the number of 1-faces of the 2-face i. N2 is the number of 2-faces of the n-polytope. For example, pn(n-cube) can be recovered from . Here, N2(n - cube) is the number of 2-faces of the n-cube: N2(n-cube) = n(n − 1)2n−3.

Moreover, we can rewrite as p. In other words, we unify into p. Although the subscript “n” is removed, the dimension n of the polytope can be determined as a result of decoding p. We stress that polytopes of different dimensions can be represented by codewords of the same format, namely two number sequences separated by “;”.

Discussion

E. A. Lazar, et al. introduced the Weinberg code to describe single Voronoi polyhedra11. But the Weingberg code does not allow for describing complexes of Voronoi polyhedra. On the other hand, our p-code allows us to describe complexes of Voronoi polyhedra, which would reveal the longer-range order of amorphous materials that cannot be seen from single Voronoi polyhedra. Our methods can be used to study a wide range of systems which are represented by polytopal tilings such as atoms in materials, grains in crystals, foams, galaxies in the universe, hyperspheres in higher-dimensional spaces, etc.1,2,3,4,5,6,7,8,9,10,11,12,21.

Conclusion

We have developed a unified theory for representing polyhedral tilings and polytopes of different dimensions by brief codewords. Specifically, we have first formulated a method to deduce how to assemble ridges to form a polyhedral tiling or a polychoron from rs = r(1)r(2)r(3)  r(R). This has been achieved by reducing the redundancy in p4. Many polychora can be constructed just from rs, but there are some polychora that need pp which contains additional information about how to assemble ridges. It is remarkable that a mere sequence of r(i)s contains all or almost all information about how to assemble r(i)-gonal ridges to form a polychoron. Since a polychoron can be constructed from , the polychoron can be represented by . The local tiling structure composed of a central polyhedron and polyhedra surrounding the central polyhedron can also be represented by , for it can be regarded as a part of a polychoron. Therefore, a polyhedral tiling can be characterized by distribution of s of different central polyhedra. The idea of assembling two-dimensional components has been generalized to higher dimensional polytopes. Using the present method, pn of an n-polytope can be converted into p whose length is as long as 1/(n − 2)! times of that of pn. Therefore, the impact of the present method factorially increases as the dimension of a polytope increases. The amount of tasks needed to convert pn to p is negligible compared to that needed to generate pn. We stress that no subscript “n” that indicates the dimension of a polytope is attached to p. The dimension of the polytope is determined as a result of decoding p. In other words, the p3-code, p4-code, p5-code, , and pn-code have been unified into the p-code. Since shorter codewords are easier to handle for both humans and computers, our unified theory of polytopes would be a powerful tool to study a wide range of structures such as atoms in materials, grains in crystals, foams, galaxies in the universe, hyperspheres in higher-dimensional spaces, etc1,2,3,4,5,6,7,8,9,10,11,12,21.

Additional Information

How to cite this article: Nishio, K. and Miyazaki, T. Describing polyhedral tilings and higher dimensional polytopes by sequence of their two-dimensional components. Sci. Rep. 7, 40269; doi: 10.1038/srep40269 (2017).

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.