DeepAI

Rounds, Color, Parity, Squares

This is a sequel to our paper "Permute, Graph, Map, Derange", involving decomposable combinatorial labeled structures in the exp-log class of type a=1/2, 1, 3/2, 2. As before, our approach is to establish how well existing theory matches experimental data and to raise open questions.

11/10/2021

Permute, Graph, Map, Derange

We study decomposable combinatorial labeled structures in the exp-log cl...
02/15/2022

Second Best, Third Worst, Fourth in Line

We investigate decomposable combinatorial labeled structures more fully,...
01/14/2022

Aspects of Muchnik's paradox in restricted betting

Muchnik's paradox says that enumerable betting strategies are not always...
04/03/2022

Combinatorial refinement on circulant graphs

The combinatorial refinement techniques have proven to be an efficient a...
05/28/2022

Overcoming Congestion in Distributed Coloring

We present a new technique to efficiently sample and communicate a large...
04/13/2020

Pentagon contact representations

Representations of planar triangulations as contact graphs of a set of i...

1 Children’s Rounds

Consider the myriad arrangements of labeled children into rounds [12, 13, 14].  A round means the same as a directed ring or circle, with exactly one child inside the ring and the others encircling.  We permit the outer ring to have as few as one child.  For , there are ways to make one round (a -round of children, inside and outside):

 4123312442133214
 4312231434122413

and ways to make two rounds (both -rounds):

 1|32|41|42|31|23|41|43|21|24|31|34|2
 2|31|42|41|33|21|43|41|24|21|34|31|2

implying that and .  For , there are ways to make a -round and ways to make a -round & a -round, implying that and .  For , there are ways to make a -round, ways to make a -round & a -round, ways to make two -rounds, and ways to make three -rounds, implying that , , , and .  We obtain

 cn=(n−1)!+(n−2)!,n≥2

and, upon normalization by ,

1000 0.621184 0.036672 0.6020 0.862134 1.317448
2000 0.622539 0.036764 0.6040 0.834706 1.312715
3000 0.623052 0.036802 0.6050 0.820866 1.311031
4000 0.623326 0.036823 0.6053 0.811867 1.310156

Table 1A: Statistics for Children’s Rounds ()

also

 limn→∞Lμnn=LG1(1,1)=0.62432998854355087099...,
 limn→∞Lσ2nn2=LG1(1,2)−LG1(1,1)2=0.03690783006485220217...,
 limn→∞Lνnn=1√e=0.60653065971263342360...,
 limn→∞Sμnln(n)=e−γ=0.56145948356688516982...,
 limn→∞Sσ2nn=SG1(1,2)=1.30720779891056809974....

It is not surprising that  enjoys linear growth:  and jointly place considerable weight on the distributional extremes.  The unusual logarithmic growth of  is due to nevertheless overwhelming all other .

1.1 Variant

If we disallow the outer ring from having just one child, then clearly , and are all zero. We obtain

 cn={0if n=2,(n−1)!+(n−2)!if n≥3
1000 0.622431 0.036818 0.6030 1.846026 3.574315
2000 0.623163 0.036837 0.6050 1.816840 3.564892
3000 0.623468 0.036851 0.6053 1.802115 3.561459
4000 0.623638 0.036860 0.6055 1.792542 3.559653

Table 1B: Statistics for Variant of Children’s Rounds ()

and, while the limits are the same as before, the limits differ by a factor of :

 limn→∞Sμnln(n)=e1−γ=1.52620511159586388047...,
 limn→∞Sσ2nn=e⋅SG1(1,2)=3.55335920579854297440...

as argued in Section 4 of [1].

2 Cycle-Colored Permutations

Assuming two colors are available [15], it is clear that

 cn=2⋅(n−1)!,n≥1.

Let us explain why and .  The permutations

 (12)(3),(13)(2)(23)(1)

each have possible colorings, giving to both and ;  the permutation contributes another to . In contrast, , and because the permutations

 (12)(34),(13)(24)(14)(23)

again give to and , whereas the permutations

 (123)(4),(132)(4),(124)(3),(142)(3),
 (134)(2),(143)(2),(234)(1),(243)(1)

give to ; also, the permutations

 (1)(2)(34),(1)(3)(24)(1)(4)(23)
 (3)(4)(12),(2)(4)(13)(2)(3)(14)

contribute another to .

Upon normalization by , we obtain

1000 0.476115 0.027160 0.4480 1.292899 1.011228
2000 0.475877 0.027132 0.4480 1.291149 0.976960
3000 0.475798 0.027123 0.4480 1.290504 0.959564
4000 0.475758 0.027119 0.4480 1.290163 0.948225

Table 2A: Statistics for Cycle-Colored Permutations ()

and

 limn→∞Lμnn=LG2(1,1)=0.47563939666525000670...,
 limn→∞Lσ2nn2=LG2(1,2)−LG2(1,1)2=0.02710536578919830440...,
 limn→∞Lνnn=−W(−e−5/4)=0.44878202648462460223...,
 limn→∞Sμn1=κ=1.29...,
 limn→∞Sσ2nln(n)=e−2γ=0.31523675168719339806....

No explicit integral is known for the constant – it appears again shortly – the limit is proved in Theorem 5 of [4].

2.1 Variant

If we prohibit -cycles, then clearly , and are all zero; however  We obtain

 cn={0if n=1,2(n−1)!if n≥2
1000 0.477065 0.027268 0.4480 3.159931 6.534345
2000 0.476353 0.027187 0.4485 3.149165 6.372009
3000 0.476115 0.027160 0.4483 3.145123 6.288169
4000 0.475996 0.027146 0.4482 3.142963 6.233112

Table 2B: Statistics for Cycle-Colored Derangements ()

and, while the limits are the same as before, the limits differ somewhat:

 limn→∞Sμn1=κ−1+e−2e−2=3.13...,
 limn→∞Sσ2nln(n)=e2−2γ=2.32930204266134332103....

The expresson involving for the average shortest cycle length follows from

 Sμn⋅bn+1⋅((n+1)!−bn)(n+1)!∼κ

and the fact that as ; hence

 Sμn⋅e−2+(1−e−2)∼κ.

Similarly,

 Sσ2n⋅bn+1⋅((n+1)!−bn)(n+1)!∼e−2γln(n)

yields

 Sσ2n⋅e−2∼e−2γln(n),

although the rightmost column of Table 2B suggests that this approximation is poor.

3 Component-Colored Mappings

Assuming three colors are available [16], it is clear that

 cn=3⋅n!n∑j=1nn−j−1(n−j)!,n≥1.

An argument in Section 3 of [1] gives , , and .  Similarly, , , , , and .

Upon normalization by , we obtain

1000 0.544944 0.032583 0.5170 2.590160 5.925381
2000 0.542744 0.032407 0.5155 2.597466 6.079830
3000 0.541778 0.032331 0.5150 2.600326 6.152861
4000 0.541205 0.032285 0.5142 2.601910 6.198088

Table 3: Statistics for Component-Colored Mappings ()

and

 limn→∞Lμnn=LG3/2(1,1)=0.53753956799272857702...,
 limn→∞Lσ2nn2=LG3/2(1,2)−LG3/2(1,1)2=0.0319941833398955610...,
 limn→∞Lνnn=sech(ξ)2=0.51092489978120431153...,
 limn→∞Sμn1=2.61...,
 limn→∞Sσ2nn1/2=6.50....

No explicit integrals are known for the latter two results.

3.1 Variant

Instead of removing the smallest possible components (-components), we wonder about removing the largest possible components.  Of course, this can be done only imprecisely, as the size of the giant component cannot be known beforehand.  If a “post-processing removal” is acceptable (rather than a “pre-processing removal”), then for the remaining components we easily have [8]

 limn→∞Lμnn=LG3/2(2,1)=0.21627840128093004373...,
 limn→∞Lσ2nn2=LG3/2(2,2)−LG3/2(2,1)2=0.00867133690820287157....

Both mean and variance are significantly reduced (from to and to , respectively).

The recursive formulas for and , however, apply expressly when the desired rank .  We cannot use our current exact integer-based algorithm to experimentally confirm these statistics for .  A Monte Carlo simulation would be feasible, but less thorough and less accurate.

4 Parity of Cycle Lengths

Let EV and OD refer to permutations with all cycle lengths even and with all cycle lengths odd, respectively

[17, 18, 19].  Clearly

 cn={(n−1)!if n≡0mod2,0if otherwise

holds for EV permutations and

 cn={(n−1)!if n≡1mod2,0if otherwise

holds for OD permutations.  On the one hand, in EV we have for all and odd . On the other hand, in OD we have always (due to the identity permutation) and , (both quickly checked).  The fact that for even but for odd gives the divergence between Tables 4B & 4C for OD.

Upon normalization by , we obtain

1000 0.758202 0.037044 0.7850 2.028405 1.400424
2000 0.758012 0.037026 0.7865 2.037816 1.400250
3000 0.757949 0.037020 0.7863 2.042013 1.400192
4000 0.757918 0.037016 0.7862 2.044523 1.400163

Table 4A: Statistics for EV permutations ()

1000 0.757601 0.036937 0.7860 0.552117 0.125460
2000 0.757712 0.036972 0.7860 0.553094 0.125440
3000 0.757749 0.036984 0.7860 0.553535 0.125434
4000 0.757768 0.036990 0.7860 0.553800 0.125431

Table 4B: Statistics for OD permutations (, )

999 0.758045 0.037077 0.7845 1.501395 1.274342
1999 0.757934 0.037042 0.7864 1.502628 1.274497
2999 0.757897 0.037031 0.7863 1.503154 1.274549
3999 0.757878 0.037025 0.7862 1.503461 1.274575

Table 4C: Statistics for OD permutations (, )

and

 limn→∞Lμnn=LG1/2(1,1)=0.75782301126849283774...,
 limn→∞Lσ2nn2=LG1/2(1,2)−LG1/2(1,1)2=0.03700721658229030320...,
 limn→∞Lνnn=4e(1+e)2=0.78644773296592741014...,
 limn→∞Sσ2nn3/2=√2SG1/2(1,2)=1.40007638550124502818...for % EV,
 limn→∞Sσ2nn3/2={0.12...for OD and n≡0mod2,1.27...for OD and n≡1mod2.

As before, no explicit integrals are known for the latter four results.

5 Square Permutations

A permutation is a square () if and only if, for any integer , the number of cycles of length (in its disjoint cycle decomposition) must be even [17, 20, 21].  There is no restriction on the number of cycles of length .  Hence both

 (1234)(5678)(9)and(123)(456)(789)

are squares, but

 (12)(34)(56)(789)and(1)(2345)(678)(9)

are not squares.  While it is not possible to enforce restrictions on -cycle counts using the values of alone, we can still employ brute force methods to calculate

 {Lk,3}3k=1={1,0,2},{Sk,3}3k=1={1,0,2}
 {Lk,4}3k=1={1,3,8},{Sk,4}2k=1={9,3};
 {Lk,5}5k=1={1,15,20,0,24},{Sk,5}5k=1={36,0,0,0,24};
 {Lk,6}5k=1={1,45,80,0,144},{Sk,6}3k=1={230,0,40};
 {Lk,7}7k=1={1,105,560,0,504,0,720},{Sk,7}7k=1={960,210,0,0,0,0,720}.

A significantly faster algorithm might provide insight on the exp-log parameter (numerical bounds, if relevant) and corresponding statistics.

We justify two of the limiting median formulas.  For ,

 1∫x1−yydy=ln(y)−y|1x=14

when , i.e., , i.e., .  For

 1∫x√1−yydy=1−x∫0√z1−zdz=−2√z+2arctanh(√z)∣∣1−x0=13

when ; this is true because

 −√1−x+arctanh(√1−x)=−tanh(ξ)+arctanh(tanh(ξ))=−tanh(ξ)+ξ=16.

7 Acknowledgements

I am grateful to Mike Spivey [22], Alois Heinz and Michael Somos for helpful discussions.  The creators of Mathematica, as well as administrators of the MIT Engaging Cluster, earn my gratitude every day.  A sequel to this paper will be released soon [23].

References

• [1] S. R. Finch, Permute, Graph, Map, Derange, arXiv:2111.05720.
• [2] S. W. Golomb and P. Gaal, On the number of permutations of objects with greatest cycle length , Adv. in Appl. Math. 20 (1998) 98–107; MR1488234.
• [3] D. Panario and B. Richmond, Exact largest and smallest size of components, Algorithmica 31 (2001) 413–432; MR1855258.
• [4] D. Panario and B. Richmond, Smallest components in decomposable structures: exp-log class, Algorithmica 29 (2001) 205–226; MR1887304.
• [5] L. A. Shepp and S. P. Lloyd, Ordered cycle lengths in a random permutation, Trans. Amer. Math. Soc. 121 (1966) 340–357; MR0195117.
• [6] T. Shi, Cycle lengths of -biased random permutations, B.S. thesis, Harvey Mudd College, 2014, http://scholarship.claremont.edu/hmc_theses/65/.
• [7] P. Flajolet and A. M. Odlyzko, Random mapping statistics, Advances in Cryptology - EUROCRYPT ’89, ed. J.-J. Quisquater and J. Vandewalle, Lect. Notes in Comp. Sci. 434, Springer-Verlag, 1990, pp. 329–354; MR1083961.
• [8] X. Gourdon, Combinatoire, Algorithmique et Géométrie des Polynômes, Ph.D. thesis, École Polytechnique, 1996.
• [9] R. Arratia, A. D. Barbour and S. Tavaré, Logarithmic Combinatorial Structures: a Probabilistic Approach, Europ. Math. Society, 2003, pp. 21-24, 52, 87–89, 118; MR2032426.
• [10] R. G. Pinsky, A view from the bridge spanning combinatorics and probability, arXiv:2105.13834.
• [11] W. J. Ewens, Population genetics theory—the past and the future, Mathematical and Statistical Developments of Evolutionary Theory, Proc. 1987 Montreal conf., Kluwer Acad. Publ., 1990, pp. 177–227; MR1108002.
• [12] R. P. Stanley, Generating functions, Studies in Combinatorics, Math. Assoc. Amer., 1978, pp. 100–141; MR0513004.
• [13] R. P. Stanley, Enumerative Combinatorics, v. 2, Cambridge Univ. Press, 1999, pp. 12–13; MR1676282.
• [14] N. J. A. Sloane, On-Line Encyclopedia of Integer Sequences, A066165, A066166, A067994, A349280, and A349426.
• [15] N. J. A. Sloane, On-Line Encyclopedia of Integer Sequences, A087981, A125553, and A349479.
• [16] N. J. A. Sloane, On-Line Encyclopedia of Integer Sequences, A063170 and A089901.
• [17] M. Bóna, A Walk through Combinatorics. An Introduction to Enumeration and Graph Theory, 4 ed., World Scientific, 2017, pp. 123–145; MR3560666.
• [18] M. Lugo, Profiles of permutations, Elec. J. Combin. 16 (2009) RP 99; MR2529808.
• [19] N. J. A. Sloane, On-Line Encyclopedia of Integer Sequences, A000246, A049218, A177145, and A204420.
• [20] H. S. Wilf, generatingfunctionology, 2 ed., Academic Press, 1994, pp. 146–150; MR1277813.
• [21] N. J. A. Sloane, On-Line Encyclopedia of Integer Sequences, A003483 and A349645.
• [22] M. Z. Spivey, A combinatorial proof for the alternating convolution of the central binomial coefficients, Amer. Math. Monthly 121 (2014) 537–540; MR3225466.
• [23] S. R. Finch, Second best, Third worst, Fourth in line, forthcoming.  Steven Finch MIT Sloan School of Management Cambridge, MA, USA steven_finch@harvard.edu