Enumeration of Mosaics / Enumeration of Canons

Harald Fripertinger

February 9, 2002

1 Enumeration of non-isomorphic mosaics

In [8] it was stated that the enumeration of mosaics is an open research problem communicated by Robert Morris from the Eastman School of Music. More information about mosaics can be found in [1]. Here I present some results from [6].

A partition p of a set X is a collection of subsets of X such that the empty set is not an element of p and such that for each x  (- X there is exactly one P  (- p with x  (- P. If p consists of exactly k subsets, then p is called a partition of size k. Let n > 1 be an integer. A partition of the set Zn := Z/nZ is called a mosaic. Let TTn denote the set of all mosaics of Zn, and let TTn,k be the set of all mosaics of Zn of size k.

A group action of a group G on the set Zn induces the following group action of G on TTn:

G × TTn -->  TTn,      (g, p) '--> gp  :=  {gP | P  (-  p},

where gP := {gi |i  (-  P }. This action can be restricted to an action of G on TTn,k. Two mosaics are called G-isomorphic if they belong to the same G-orbit on TTn, in other words, p1,p2  (- TTn are isomorphic if gp1 = p2 for some g  (- G. Usually the cyclic group Cn := <(0, 1,...,n- 1)>, the dihedral group Dn := <(0, 1,...,n- 1), (0,n- 1)(1,n- 2)...>, or the group of all affine mappings on Zn are candidates for the group G. If G acts on a set X, then for each g  (- G the permutation x'-->gx is indicated by g . It is called the permutation representation of g. A detailed introduction to combinatorics under finite group actions can be found in [910].

It is well known (see [45]) how to enumerate G-isomorphism classes of mosaics (i.e. G-orbits of partitions) by identifying them with G×Sn-orbits on the set of all functions from Zn to n := {1,...,n}. (The symmetric group of the set n is denoted by Sn.) Furthermore, G-mosaics of size k correspond to G×Sk-orbits on the set of all surjective functions from Zn to k.

Theorem 1. Let Mk be the number of G × Sk-orbits on kZn, then the number of G-isomorphism classes of mosaics of Zn is given by Mn, and the number of G-isomorphism classes of mosaics of size k is given by Mk - Mk-1, where M0 := 0.

Using the Cauchy-Frobenius-Lemma, we have

                         n
       ---1----   sum       prod      i ai(g)
Mk  =  |G ||S |             a1(s )   ,
            k (g,s) (- G×Sk i=1

where ai(g ) or ai(s) are the numbers of i-cycles in the cycle decomposition of g or s respectively.

Finally, the number of G-isomorphism classes of mosaics of size k can also be derived by the Cauchy-Frobenius-Lemma for surjective functions by

                  c(s)              k  (     )  n  (         )aj(g)
---1----    sum      sum       c(s)- l sum   prod    ai(s)   prod    s um 
|G ||S  |             (-1)                a             d .ad      ,
      k (g,s) (- G×Sk l=1           a  i=1     i   j=1   d|j

where the inner sum is taken over the sequences a = (a1,...,ak) of nonnegative integers ai such that  sum i=1ka i = l, and where c(s) is the number of all cycles in the cycle decomposition of s.

If p  (- TTn consists of ci blocks of size i for i  (- n , then p is said to be of block-type c = (c1,...,cn). From the definition it is obvious that  sum i=1nic i = n, which will be indicated by c  |-  -| n. Furthermore, it is clear that p is a partition of size  sum i=1nc i. The set of mosaics of block-type c will be indicated as TTc. Since the action of G on TTn can be restricted to an action of G on TTc, we want to determine the number of G-isomorphism classes of mosaics of type c. For doing that, let c be any partition of type c. (For instance, c can be defined such that the blocks of c of size 1 are given by {1}, {2}, ..., {c1}, the blocks of c of size 2 are given by {c1 + 1,c1 + 2}, {c1 + 3,c1 + 4}, ..., {c1 + 2c2 -  1,c1 + 2c2}, and so on.) According to [910], the stabilizer Hc of c in the symmetric group Sn is similar to the direct sum

 o+  n
    S  [S ]
      ci i
 i=1

of compositions of symmetric groups, which is a permutation representation of the direct product

 n
×i=1 Si|Sci-

of wreath products of symmetric groups. In other words Hc is the set of all permutations s  (- Sn, which map each block of the partition c again onto a block (of the same size) of the partition.

Hence, the G-isomorphism classes of mosaics of type c can be described as G × Hc-orbits of bijections from Zn to n under the following group action:

             Z      Z                            -1
(G  × Hc) ×  nbnij -->  nbnij,     ((g, s),f) '-->  g o f o s .

When interpreting the bijections from Zn to n as permutations of the n-set n, then G-mosaics of type c correspond to double cosets (cf. [910]) of the form

G\Sn /Hc.
    --

Theorem 2. The number of G-isomorphism classes of mosaics of type c is given by

                  n
---1----    sum      prod   a (s)!iai(g),
|G ||Hc |              i
         (g,sz()g (- )=Gz×(sH)c i=1

where z(g ) and z(s) are the cycle types of g and of s respectively, given in the form (ai(g ))i (- n or (ai(s))i (- n. In other words we are summing over those pairs (g,s) such that g and s determine permutations of the same cycle type.

2 Enumeration of non-isomorphic canons

The present concept of a canon is described by G. Mazzola in [11] and was presented by him to the author in the following way: A canon is a subset K  (_ Zn together with a covering of K by pairwise different subsets V i/=Ø for 1 < i < t, the voices, where t > 1 is the number of voices of K, in other words

      t
      U 
K  =    Vi,
     i=1

such that for all i,j  (- {1, ...,t}

  1. the set V i can be obtained from V j by a translation of Zn,
  2. there is only the identity translation which maps V i to V i,
  3. the set of differences in K generates Zn, i.e. <K - K> := <k - l | k,l  (- K> = Zn.

We prefer to write a canon K as a set of its subsets V i. Two canons K = {V1,...,Vt} and L = {W1, ...,Ws} are called isomorphic if s = t and if there exists a translation T of Zn and a permutation p in the symmetric group St such that T(V i) = Wp(i) for 1 < i < t. Then obviously T(K) = L.

Here we present some results from [7]. The cyclic group Cn acts on the set of all functions from Zn to {0,1} by

           Zn         Zn                   -1
Cn × {0,1}   -- >  {0, 1}       (s,f) '-->  f o s  .

As the canonical representative of the orbit Cn(f) = {f o s |s  (-  Cn} we choose the function f0  (- Cn(f), such that f0 < h for all h  (- Cn(f).

A function f  (- {0, 1}Zn (or the corresponding vector (f(0),f(1),...,f(n - 1))) is called acyclic if Cn(f) consists of n different objects. The canonical representative of the orbit of an acyclic function is called a Lyndon word.

As usual, we identify a subset A of Zn with its characteristic function xA : Zn -->{0,1} given by

        { 1   if i  (-  A
xA(i) =
          0   otherwise.

Following the ideas of [7] and the notion of [3], a canon can be described as a pair (L,A), where L is the inner and A the outer rhythm of the canon. In other words, the rhythm of one voice is described by L and the distribution of the different voices is described by A, i.e. the onsets of the different voices are a + L for a  (- A. In the present situation, L/=0 is a Lyndon word of length n over the alphabet {0,1}, and A is a t-subset of Zn. But not each pair (L,A) describes a canon. More precisely we have:

Lemma 3. The pair (L,A) does not describe a canon in Zn if and only if there exists a divisor d > 1 of n such that L(i) = 1 implies i  =_ d - 1 mod d and xA0(i) = 1 implies i  =_ d - 1 mod d, where xA0 is the canonical representative of Cn(xA).

An application of the principle of inclusion and exclusion allows to enumerate the number of non isomorphic canons.

Theorem 4. The number of isomorphism classes of canons in Zn is

        sum 
Kn  :=     m(d)c(n/d)a(n/d),
       d|n

where m is the Moebius function, c(1) = 1,

           sum 
c(r) =  1-   m(s)2r/s for r > 1,
        r s|r

and

           sum 
a(r) =  1-   f(s)2r/s-  1 for r > 1,
        r s|r

where f is the Euler totient function.

3 Enumeration of rhythmic tiling canons

There exist more complicated definitions of canons. A canon described by the pair (R,A) of inner and outer rhythm defines a rhythmic tiling canon in Zn with voices V a for a  (- A if and only if

  1. the voices V a cover entirely the cyclic group Zn,
  2. the voices V a are pairwise disjoint.

Rhythmic tiling canons with the additional property

  1. both R and A are aperiodic,

are called regular complementary canons of maximal category.

Hence, rhythmic tiling canons are canons which are also mosaics. More precisely, if |A | = t, then they are mosaics consisting of t blocks of size n/t, whence they are of block-type c where

     {
ci =   t  if i = n/t
       0  otherwise.

This block-type will be also indicated as c = ((n/t)t). So far the author did not find a characterization of those mosaics of block-type c describing canons, which could be used in order to apply enumeration formulae similar to those for the enumeration of non-isomorphic mosaics. Applying Theorem 2 the numbers of Cn-isomorphism classes of mosaics presented in table 1 were computed.


    |
  n |
-12-|(62) :-44-(43) :-499-(34)-: 1306--(26) : 902---------------------------------
    |   2            3                4                   6
 24 |(12 ) : 56450 (8 ) : 65735799  (6 ) : 4008.268588  (4 ) : 187886.308429
    |(38) : 381736.855102  (212) : 13176.573910
 36 |(182) : 126047906  (123) : 15.670055.601970  (94) : 24829.574426.591236
    |(66) : 103.016116.387908.956698  (49) : 10778.751016.666506.604919
    |  12                                 18
    |(3 2) : 9910.160306.188702.4944292  (2  ) : 6.156752.656678.674792
 40 |(20 ) : 1723.097066  (10 ) : 4.901417.574950.588294
    |(85) : 1595.148844.422078.211829  (58) : 11.765613.697294.131102.617360
    |(410) : 88.656304.986604.408738.684375  (220) : 7995.774669.504366.055054


Table 1: Number of mosaics in Zn of block-type ((n/t)t)

However, the description of the isomorphism classes of canons as pairs (L,Cn(A)) consisting of Lyndon words L and Cn-orbits of subsets A of Zn with some additional properties (c.f. Lemma 3) can also be applied for the determination of complete sets of representatives of non-isomorphic canons in Zn, as was indicated in the last part of [7].

There exist fast algorithms for computing all Lyndon words of length n over {0,1} and all Cn-orbit representatives of subsets of Zn. For finding regular tiling canons with t voices (where t is necessarily a divisor of n), we can restrict to Lyndon words L with exactly n/t entries 1 and to representatives A0 of the Cn-orbits of t-subsets of Zn. Then each pair (L,A0) must be tested whether it is a regular tiling canon. In this test we only have to test whether the voices described by (L,A0) determine a partition on Zn, because in this case it is obvious that (L,A0) does not satisfy the assumptions of Lemma 3.

For finding the number of regular tiling canons, we make use of still another result. In [132] it is shown that regular complementary canons of maximal category occur only for certain values of n, actually only for non-Hajós-groups Zn (cf. [12]). The smallest n for which Zn is not a Hajós-group is n = 72 which is still much further than the scope of our computations. Hence, we deduce that for all n such that Zn is a Hajós-group the following is true:

Lemma 5. If a pair (L,A0) describes a regular tiling canon in a Hajós-group Zn, then A0 is not aperiodic.

This reduces dramatically the number of pairs which must be tested.

By applying Theorem 4 we computed Kn, the numbers of non-isomorphic canons of length n, given in the third column of table 2. The construction described above yields a list of all regular tiling canons, which also yields Tn, the number of regular tiling canons of length n, given in the second column of table 2.


n Tn Kn



2 1 1
3 1 5
4 2 13
5 1 41
6 3 110
7 1 341
8 6 1035
9 4 3298
10 6 10550
11 1 34781
12 23 117455
13 1 397529
14 13 1.370798
15 25 4.780715
16 49 16788150
17 1 59451809
18 91 212.178317
19 1 761.456429
20 149 2749.100993
21 121 9973.716835
22 99 36347.760182
23 1 133022.502005
24 794 488685.427750
25 126 1.801445.810166
26 322 6.662133.496934
27 766 24.711213.822232
28 1301 91.910318.016551
29 1 342.723412.096889
30 3952 1281.025524.753966
31 1 4798.840870.353221
32 4641 18014.401038.596400
33 5409 67756.652509.423763
34 3864 255318.257892.932894
35 2713 963748.277489.391403
3631651 3.643801.587330.857840
37 1 13.798002.875101.582409
3813807 52.325390.403899.973926
3940937198.705759.014912.561995
4064989755.578639.350274.265100
Table 2: Number of non-isomorphic canons in Zn

The group Zn is a Hajós group if the decomposition of n is not too complicated. If n is of the form
pk for k > 0, pkq for k > 1, p2q2, pkqr for k  (-  {1,2}, pqrs

for distinct primes p, q, r and s, then Zn is a Hajós group and Vuza proved in [1314] that for these n there do not exist regular complementary canons of maximal category. Moreover, he described a method how to construct such canons for all Zn which are not Hajós groups. If Zn is not a Hajós group, then n can be expressed in the form p1p2n1n2n3 with p1, p2 primes, ni > 2 for 1 < i < 3, and gcd(n1p1,n2p2) = 1. Vuza presents an algorithm for constructing two aperiodic subsets L and A of Zn, such that |L | = n1n2, |A | = p1p2n3, and L + A = Zn. Hence, L or A can serve as the inner rhythm and the other set as the outer rhythm of such a canon. Moreover, it is important to mention that there is some freedom for constructing these two sets, and each of these two sets can be constructed independently from the other one. He also proves that when L and A satisfy L + A = Zn, then also (kL,A), (kL,kA) have this property for all k  (- Zn*.

So the first step in enumeration of regular complementary canons of maximal category is to determine the number of non-isomorphic canons which can be constructed by this method. (Actually, I wanted to do it this week, but finally there was not enough time to do so.)

4 Some interesting open problems

  1. In his papers, Vuza did not prove that each regular complementary canon of maximal category can be constructed with his method. Is it possible to find regular complementary canons of maximal category which cannot be produced by Vuza’s approach?
  2. Is there a more elegant method for enumerating regular tiling canons?
  3. When enumerating isomorphism classes of mosaics in Zn, we could apply groups different from the cyclic group Cn. How to do this for canons? For the group Cn, a canon was given as a pair (L,A) with certain properties. L was an acyclic vector, so probably in all generalizations we must assume that L does not have cyclic symmetries. A was considered to be a subset of Zn, but actually A describes the onset distribution of the different voices, whence it is actually a subset of the acting group Cn. When considering the group Aff 1(Zn), consisting of all affine mappings pa,b : Zn --> Zn, i'-->ai + b, for a  (- Zn* and b  (- Z n, acting on Zn, then A must be considered as a subset of this group. If L has just the trivial symmetry, then each pa,b(L) describes another voice of the canon. If the stabilizer U of L is non-trivial, but it does not contain symmetries of the form p1,b for b/=0, then A must be considered as a subset of the right-cosets of U in Aff 1(Zn). When computing the number of non-isomorphic canons in this setting, we get a much bigger number of different canons, since usually many different voices start at the same onset. So maybe in this situation we should restrict to canons, such that different voices have different onsets in Zn. But when speaking of onsets of voices we can get some problems with symmetries pa,b of L for a/=1 and b/=0. So maybe we should not allow any symmetries of L. But then we will not get a complete overview over all canons in Zn. Still the property that K - K generates Zn was not considered for these generalizations.

References

[1]   B. Alegant. The Seventy-Seven Partitions of the Aggregate: Analytical and Theoretical Implications. PhD thesis, Eastman School of Music, University of Rochester, 1992.

[2]   M. Andreatta. Group theoretical methods applied to music. Visiting Student Dissertation, University of Sussex, 1997.

[3]   M. Andreatta, Th. Noll, C. Agon, and G. Assayag. The geometrical groove: rhythmic canons between theory, implementation and musical experiment. In Les Actes des 8e Journées d’Informatique Musicale, Bourges 7-9 juin 2001, pages 93-97, 2001.

[4]   N.G. de Bruijn. Pólya’s Theory of Counting. In E.F. Beckenbach, editor, Applied Combinatorial Mathematics, chapter 5, pages 144 - 184. Wiley, New York, 1964.

[5]   N.G. de Bruijn. On the number of partition patterns of a set. Nederl. Akad. Wetensch. Proc. Ser. A 82 = Indag. Math. 41, pages 229 - 234, 1979.

[6]   H. Fripertinger. Enumeration of Mosaics. Discrete Mathematics, 199:49-60, 1999.

[7]   H. Fripertinger. Enumeration of non-isomorphic canons. To appear in the Tatra Mountains Mathematical Publications journal, 2001.

[8]   P. Isihara and M. Knapp. Basic Z12-Analysis of Musical Chords. UMAP Journal 14.4, pages 319 - 348, 1993.

[9]   A. Kerber. Algebraic Combinatorics via Finite Group Actions. B.I. Wissenschaftsverlag, Mannheim, Wien, Zürich, 1991. ISBN 3-411-14521-8.

[10]   A. Kerber. Applied Finite Group Actions, volume 19 of Algorithms and Combinatorics. Springer, Berlin, Heidelberg, New York, 1999. ISBN 3-540-65941-2.

[11]   G. Mazzola. Topos of music. To appear.

[12]   A. Sands. The factorization of Abelian Groups (II). Quart. J. Math. Oxford, 10(2):45-54, 1962.

[13]   D.T. Vuza. Supplementary sets and regular complementary unending canons (part one). Perspectives of New Music, 29(2):22-49, 1991.

[14]   D.T. Vuza. Supplementary sets and regular complementary unending canons (part four). Perspectives of New Music, 31(1):270-305, 1993.

Address:
Harald Fripertinger
Institut für Mathematik
Karl-Franzens-Universität Graz
Heinrichstr. 36/4
A-8010 Graz
Austria
harald.fripertinger@uni-graz.at