Loading [MathJax]/jax/output/SVG/jax.js

On the existence of permutations conditioned by certain rational functions

  • Received: 01 October 2019 Revised: 01 February 2020
  • 05A05, 05B99, 05C05

  • We prove several conjectures made by Z.-W. Sun on the existence of permutations conditioned by certain rational functions. Furthermore, we fully characterize all integer values of the "inverse difference" rational function. Our proofs consist of both investigation of the mathematical properties of the rational functions and brute-force attack by computer for finding special permutations.

    Citation: Guo-Niu Han. On the existence of permutations conditioned by certain rational functions[J]. Electronic Research Archive, 2020, 28(1): 149-156. doi: 10.3934/era.2020009

    Related Papers:

    [1] Guo-Niu Han . On the existence of permutations conditioned by certain rational functions. Electronic Research Archive, 2020, 28(1): 149-156. doi: 10.3934/era.2020009
    [2] Bin Han . Some multivariate polynomials for doubled permutations. Electronic Research Archive, 2021, 29(2): 1925-1944. doi: 10.3934/era.2020098
    [3] Juan Gerardo Alcázar, Carlos Hermoso, Hüsnü Anıl Çoban, Uğur Gözütok . Computation of symmetries of rational surfaces. Electronic Research Archive, 2024, 32(11): 6087-6108. doi: 10.3934/era.2024282
    [4] Heesung Shin, Jiang Zeng . More bijections for Entringer and Arnold families. Electronic Research Archive, 2021, 29(2): 2167-2185. doi: 10.3934/era.2020111
    [5] Kelvyn Bladen, D. Richard Cutler . Assessing agreement between permutation and dropout variable importance methods for regression and random forest models. Electronic Research Archive, 2024, 32(7): 4495-4514. doi: 10.3934/era.2024203
    [6] Shishuo Fu, Zhicong Lin, Yaling Wang . Refined Wilf-equivalences by Comtet statistics. Electronic Research Archive, 2021, 29(5): 2877-2913. doi: 10.3934/era.2021018
    [7] Jin Li, Yongling Cheng . Barycentric rational interpolation method for solving time-dependent fractional convection-diffusion equation. Electronic Research Archive, 2023, 31(7): 4034-4056. doi: 10.3934/era.2023205
    [8] Jin Li, Yongling Cheng . Barycentric rational interpolation method for solving fractional cable equation. Electronic Research Archive, 2023, 31(6): 3649-3665. doi: 10.3934/era.2023185
    [9] Hai-Liang Wu, Li-Yuan Wang . Permutations involving squares in finite fields. Electronic Research Archive, 2022, 30(6): 2109-2120. doi: 10.3934/era.2022106
    [10] Caiwen Chen, Tianxiu Lu, Ping Gao . Chaotic performance and circuitry implement of piecewise logistic-like mapping. Electronic Research Archive, 2025, 33(1): 102-120. doi: 10.3934/era.2025006
  • We prove several conjectures made by Z.-W. Sun on the existence of permutations conditioned by certain rational functions. Furthermore, we fully characterize all integer values of the "inverse difference" rational function. Our proofs consist of both investigation of the mathematical properties of the rational functions and brute-force attack by computer for finding special permutations.



    Permutations (see, for example, [1,3]) are studied in almost every branch of mathematics and also in computer science. The number of permutations π=(π(1),π(2),,π(n))Sn of {1,2,,n} is n!. In [4] Z.-W. Sun made several conjectures about the existence of permutations conditioned by certain rational functions. In the paper we confirm three of them by proving the following theorem.

    Theorem 1. (ⅰ) For any integer n>5, there is a permutation πSn such that

    n1k=11π(k)π(k+1)=0. (1.1)

    (ⅱ) For any integer n>7, there is a permutation πSn such that

    n1k=11π(k)π(k+1)+1π(n)π(1)=0. (1.2)

    (ⅲ) For any integer n>5, there is a permutation πSn such that

    n1k=11π(k)π(k+1)=1. (1.3)

    Since n! is a huge number for large n, the generation of all n! permutations of n by computer is already a challenge [2]. For this reason (1.1)-(1.3) have only been verified for very small n. Our proof of Theorem 1 consists of both investigation of the mathematical properties of the rational functions and brute-force attack by computer for finding certain special permutations.

    Furthermore, we can fully characterize all integer values of the "inverse difference" rational function given in the left-hand side of (1.1). We define

    Vn={n1k=11π(k)π(k+1):πSn}. (1.4)

    Since aVn implies aVn by the reverse of permutation, we only need to study the nonnegative integer values of Vn. For example, n=5, the value set V5 contains the following nonnegative rational numbers:

    112,16,14,13,12,712,23,34,1,76,43,32,1912,74,116,2312,2,136,114,4.

    We see that there are three integers 1,2,4 in the above list.

    Theorem 2. We have V3N={2}, V5N={1,2,4}, and for n3,5,

    VnN={0jn1jn2}. (1.5)

    The proofs of Theorems 1 and 2 will be given in Section 2. Notice that we are still not able to prove three other conjectures of Sun. Let us reproduce them below for interested readers.

    Conjecture 3. (i) [4,Conj. 4.7(ⅱ)] For any integer n>6, there is a permutation πSn such that

    n1k=11π(k)+π(k+1)=1. (1.6)

    Also, for any integer n>7, there is a permutation πSn such that

    n1k=11π(k)+π(k+1)+1π(n)+π(1)=1. (1.7)

    (ii) [4,Conj. 4.8(ⅱ)] For any integer n>7, there is a permutation πSn such that

    n1k=11π(k)2π(k+1)2=0. (1.8)

    Motivated by (1.8), we make the following conjecture.

    Conjecture 4. For any integer n>11, there is a permutation πSn such that

    n1k=11π(k)2π(k+1)2+1π(n)2π(1)2=0. (1.9)

    Conjecture 4 has been checked for 11<n<28 by computer. We list below the permutations satisfying (1.9), which are found by our computer program in a highly non-trivial way.

    π12=(1,4,3,5,7,2,12,8,10,11,9,6),π13=(1,2,12,8,9,6,11,10,7,5,13,4,3),π14=(1,2,12,9,6,4,3,13,8,7,5,10,14,11),π15=(1,9,2,3,12,10,11,5,4,14,6,15,13,8,7),π16=(1,3,2,4,5,11,16,14,10,8,6,12,9,15,13,7),π17=(1,3,2,4,5,9,15,6,12,16,11,10,14,13,8,7,17),π18=(1,3,2,4,6,5,7,13,8,14,12,16,10,18,17,9,11,15),π19=(1,3,2,4,6,5,7,8,12,18,17,13,9,15,11,10,16,19,14),π20=(1,3,2,4,6,5,7,18,8,13,12,17,9,20,16,19,10,11,15,14),π21=(1,3,2,4,6,5,7,17,8,20,16,9,12,18,15,13,19,21,11,14,10),π22=(1,3,2,4,6,5,7,8,20,13,17,22,18,12,9,15,21,19,16,11,10,14),π23=(1,3,2,4,6,14,10,18,12,8,20,7,5,21,15,11,17,13,22,23,16,19,9),π24=(1,3,2,4,6,14,10,18,12,8,5,9,21,11,24,16,20,22,17,15,13,19,23,7),π25=(1,3,2,4,6,14,10,18,12,8,5,16,24,9,21,23,7,17,15,11,13,22,20,19,25),π26=(1,3,2,4,6,14,10,18,12,8,22,13,5,23,16,20,19,21,9,7,17,11,25,15,24,26),π27=(1,3,2,4,6,14,10,18,12,8,22,13,9,5,11,21,23,16,26,19,25,27,17,15,24,20,7).

    Let Φdif(π), Φcycdif(π), and Φprod(π) denote the three rational functions expressed in the left-hand side of (1.1), (1.2), and (1.3), respectively. The following Link lemma is useful for our construction.

    Lemma 5 (Link). Let σSs and τSt be two permutations on {1,2,,s} and {1,2,,t}, respectively, such that σ(s)=s,τ(1)=1 and Φdif(σ)=Φdif(τ)=0. We define the "link" of the two permutations ρSs+t1 by

    ρ(k)={σ(k),if 1kss1+τ(ks+1).if s+1ks+t1

    Then, we have Φdif(ρ)=0. Furthermore, if τ(t)=t, we have ρ(s+t1)=s+t1.

    Proof. Notice that in the definition of ρ, if we allow k=s in the second case, the expression will give the same definition of ρ(s) as in the first case, since σ(s)=s=s1+τ(1). Hence,

    Φdif(ρ)=s1k=11ρ(k)ρ(k+1)+s+t2k=s1ρ(k)ρ(k+1)=s1k=11σ(k)σ(k+1)+t1k=11τ(k)τ(k+1)=Φdif(σ)+Φdif(τ)=0.

    Furthermore, if τ(t)=t, it is easy to see that ρ(s+t1)=s+t1.

    Let us write the link ρ of σ and τ by σ,τ.

    Example. Take σ=(1,4,2,5,3,6) and τ=(1,3,2,4). We verify that

    Φdif(σ)=13+1213+1213=0,Φdif(τ)=12+1112=0.

    We have ρ=σ,τ=(1,4,2,5,3,6,8,7,9) and

    Φdif(ρ)=(13+1213+1213)+(12+1112)=0.

    If τ(t)=t, since the link ρ=σ,τ also satisfies the conditions ρ(s+t1)=s+t1 and Φ(ρ)=0, we can "link" again and obtain ρ,τ=σ,τ,τ.

    The following proposition is a slightly stronger version of Theorem 1(ⅰ) for the rational function Φdif.

    Proposition 6. For any integer n>5, there is a permutation πSn such that π(1)=1,π(n)=n, and Φdif(π)=0.

    Proof. Let

    σ0=(1,4,2,5,3,6),σ1=(1,3,2,4),σ2=(1,3,6,4,7,5,2,8),

    and τ=σ1=(1,3,2,4). We have Φdif(σj)=0 for j=0,1,2, and τ(1)=1,τ(4)=4. By repeated application of the link algorithm, we obtain the following three families of permutations:

    (1,4,2,5,3,6),(1,4,2,5,3,68,7,9),(1,4,2,5,3,68,7,911,10,12),(1,3,2,4),(1,3,2,46,5,7),(1,3,2,46,5,79,8,10),(1,3,6,4,7,5,2,8),(1,3,6,4,7,5,2,810,9,11),(1,3,6,4,7,5,2,810,9,1113,12,14),

    of length

    n=6,9,12,15,(3k)n=4,7,10,13,(3k+1)n=8,11,14,17(3k+2)

    Hence we have constructed one permutation πSn for each n>5 such that Φdif(π)=0 and π(1)=1 and π(n)=n.

    The following proposition is another enhanced version of Theorem 1(ⅰ) for Φdif, which is crucial for proving Theorem 1(ⅱ) for Φcycdif. The difference between Propositions 6 and 7 lies in the value of π(n).

    Proposition 7. For any integer n>7, there is a permutation πSn such that π(1)=1,π(n)=n1 and Φdif(π)=0.

    Proof. Let

    α8=(1,2,4,8,6,5,3,7),α9=(1,4,2,5,9,3,7,6,8),α10=(1,2,6,3,7,8,5,4,10,9),α11=(1,2,3,4,6,5,9,8,7,11,10),α12=(1,2,3,6,4,8,12,10,9,7,5,11).

    We have αj(1)=1, αj(j)=j1, and Φdif(αj)=0 for j=8,9,,12. The proposition is true for j=8,9,,12. For n13 and k=n76, take the permutation σSk obtained in Proposition 6, i.e., σ(1)=1,σ(k)=k,Φdif(σ)=0. Then, the link ρ=σ,α8 satisfies ρ(1)=1,ρ(n)=n1 and Φdif(ρ)=0. For example, for n=13 and k=6,

    ρ=(1,4,2,5,3,6),(1,2,4,8,6,5,3,7)=(1,4,2,5,3,6,7,9,13,11,10,8,12).

    Hence, the proposition is true for any n>7.

    Now we are ready to prove part (ⅱ) of Theorem 1 for the rational function Φcycdif.

    Proof of Theorem 1(ii). If n=2k is even, we can easily check that the permutation

    π=(1,2,3,,k1,k,2k,2k1,,k+3,k+2,k+1),

    which is obtained by the concatenation of the increasing permutation of Sk and the decreasing permutation of {jk+1j2k}, satisfies

    Φcycdif(π)=(1111111k+11+11++11)+1k=0.

    The odd case is more complicated. First, we define

    β9=(2,1,4,5,9,3,7,6,8),β11=(1,2,11,5,4,8,7,9,3,6,10),β13=(1,2,13,3,5,4,9,8,10,6,11,7,12).

    We have Φcycdif(βj)=0 for j=9,11,13. Next, for n=2k+115, i.e., k7 and m=k+18, by Propositions 6 and 7, there exist two permutations σSk and σSm such that

    (ⅰ) σ(1)=1, σ(k)=k, and Φdif(σ)=0;

    (ⅱ) τ(1)=1, τ(m)=m1, and Φdif(τ)=0.

    Let τ be the permutation of {jk+1j2k+1} obtained by adding k in the reverse of τ:

    τ=(k+τ(m),k+τ(m1),k+,k+τ(3),k+τ(2),k+τ(1)).

    Notice that the first and last elements of τ are k+τ(m)=2k and k+τ(1)=k+1, respectively. Let ρ be the concatenation of σ and τ. We can verity that Φdif(τ)=Φdif(τ)=0 and

    Φcycdif(ρ)=Φdif(σ)+1k2k+Φdif(τ)+1(k+1)1=0.

    For example, for n=15, k=7 and m=8, we have σ=(1,3,2,4,6,5,7) and τ=(1,2,4,8,6,5,3,7), so that τ=(14,10,12,13,15,11,9,8). Our final permutation ρ is the concatenation of σ and τ:

    ρ=(1,3,2,4,6,5,7, 14,10,12,13,15,11,9,8).

    We can check that Φcycdif(ρ)=0.

    To prove part (ⅲ) of Theorem 1 concerning the rational function Φprod, we need the following lemma. Also, it is much more convenient to describe the construction in the increasing binary trees model (see, for example, [3,p. 51], [1,p. 143]).

    Lemma 8 (Insertion). Let σ=σ(1)σ(2)σ(n1)Sn1 be a permutation and τSn be the permutation obtained by insertion of the letter n into σ:

    τ=σ(1)σ(j)nσ(j+1)σ(n1).(j=1,2,,n2)

    Then, Φprod(σ)=Φprod(τ) if and only if σ(j)+σ(j+1)=n.

    Proof. We have

    Φprod(σ)=n2k=11σ(k)σ(k+1)=+1σ(j)σ(j+1)+Φprod(τ)=n1k=11τ(k)τ(k+1)=+1σ(j)n+1nσ(j+1)+=+σ(j)+σ(j+1)nσ(j)σ(j+1)+

    Hence, Φprod(σ)=Φprod(τ) if and only if

    1σ(j)σ(j+1)=σ(j)+σ(j+1)nσ(j)σ(j+1),

    i.e., σ(j)+σ(j+1)=n.

    Now we use the insertion lemma to prove part (ⅲ) of our main theorem.

    Proof of Theorem 1(iii). Let

    δ6=(2,1,3,4,5,6),δ7=(2,1,3,7,4,5,6),δ8=(6,4,1,2,7,5,3,8).

    We verify that Φprod(δj)=1 for j=6,7,8. By insertion lemma, we define σ9 by inserting the letter 9 between 2 and 7 in σ8:

    σ9=(6,4,1,2,9,7,5,3,8).

    Next, we insert 10 in σ9 between 6 and 4:

    σ10=(6,10,4,1,2,9,7,5,3,8).

    For the insertion of 11, we have two possible positions, namely, between (2,9) and (3,8). We choose the position (2,9) and define

    σ11=(6,10,4,1,2,11,9,7,5,3,8).

    The crucial idea is to show that this kind of insertion can be repeatedly applied as many times as we want, starting from σ8. Thus, we obtain the desired permutation σnSn for each n8. To understand the general pattern, we take a rather big example with n=32. Our permutation δ32 is as follows:

    δ32=(6,16,10,24,14,32,18,22,26,30,4,1,2,31,29,27,25,23,21,19,17,15,28,13,11,20,9,7,12,5,3,8),

    which can be represented by the increasing binary tree (see, for example, [3,p. 51], [1,p. 143]) in Figure 1. We see that the nodes of the form 2k+1,4k+2,8k,8k+4 all appear in the tree structure. Hence, the insertion can be repeatedly applied to reach each δn for n8.

    Figure 1.  The increasing binary tree for δ32.

    As proved in Theorem 1, for any integer n>5, there is a permutation πSn such that Φdif(π)=0. For a permutation π, the value of Φdif(π) is, a priori, a rational number. Theorem 2 provides a full characterization of the integer values of the rational function Φdif.

    Proof of Theorem 2. In fact, we need to prove a stronger statement by replacing each Vn in the theorem by

    Vn={n1k=11π(k)π(k+1):πSn,π(n)=n}. (2.1)

    It is easy to see that n1Vn by taking the identity permutation (1,2,n). Also, the maximal value of Φdif(π) is n1, so that mVn for mn. We can check by computer for all n6. If n7, we prove by induction on n. First, 0Vn by Proposition 6. Next, for 1mn3, by the induction hypothesis, there is a permutation τSn1 such that τ(n1)=n1 and Φdif(τ)=m1. We define

    π=(τ(1),τ(2),,τ(n1),n).

    We verify that Φdif(π)=1+Φdif(τ)=m. Finally, for each σSn which is not the identity permutation, we see that there exists at least one negative term in the summation (1.1). Hence Φdif(σ)<n2, and n2Vn.

    For n=1,2, and k=0,1,,n1, let η(n,k) be the number of permutations πSn satisfying

    n1k=11π(k)π(k+1)=k.

    We list the first values of η(n,k) in the following table.

    nk012345678910111120130014220150340161216360171844388801834833913673111001990612848022641121812011097408112408119744481812714011140992444622802810695375484221214160112376976340032178236810372526665481272288241801


    [1] (2009) Analytic Combinatorics. Cambridge: Cambridge University Press.
    [2] Permutation generation methods. Comput. Surveys (1977) 9: 137-164.
    [3] (2012) Enumerative Combinatorics. Cambridge: Cambridge Studies in Advanced Mathematics, 49, Cambridge University Press.
    [4] Z.-W. Sun, On permutation of {1,,n} and related topics, preprint, arXiv: 1811.10503.
  • This article has been cited by:

    1. Zhi-Wei Sun, On permutations of $$\{1,\ldots ,n\}$$ and related topics, 2021, 54, 0925-9899, 893, 10.1007/s10801-021-01028-8
  • Reader Comments
  • © 2020 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(2553) PDF downloads(168) Cited by(1)

Article outline

Figures and Tables

Figures(1)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog