-1

二分完美洗牌排列 将$i \rightarrow 2i( mod (2n+1))$.

$2n+1=3^k$时,很容易得出2分完美洗牌中的$k$个循环结构体,首先我们需要以下定理:

定理1 当p为奇素数且$g$$p^2$的原根时$\Rightarrow$ $g$也是$p^k$的原根

定义1 欧拉函数,$\phi(m)$定义为小于等于$m$中,与$m$互素的正整数个数

定义2 对模指数,$a$对模$m$的对模指数定义为 $Ord_m(a)=d, st: a^d\equiv1(mod m)$中最小的d

定义3$\phi(m)=Ord_m(a)$ 则称$a$$m$的原根

性质1$(a,m)=1$$a$$m$的原根,那么$a$$(\Bbb Z/n\Bbb Z)^*$的生成元

例如n=4,则$2n+1=3^2$, ${1,2,3,4,5,6,7,8,9}包含3个循环结构{1,2,4,5,7,8} {3,6}以及{9}

与9互素的正整数集合为{1,2,4,5,7,8},因此$\phi(m)=6$, 而满足$2^d\equiv1 (mod9)$的最小$d=6$ 因此2为9的原根,同时结合原根的性质2也是整数模9乘法群${1,2,4,5,7,8}$的生成元

考虑包含$3^k$因子的其他循环结构,这些循环结构的元素必定是$3^s2^t$形式 //问题1 如何证明?

当s=0时,循环结构为{1,2,4,5,7,8}

当s=1时,循环结构为{3,6}

当s=2时,循环结构为{9}

问题2,这些结构并非循环群,为何仍然使用生成元,是否正确?

===========================================================

consider the 2 way perfect shuffle with n=4, that is with elements 1 to 8, how to prove that there are exactly 3 cyclic structures within the shuffle array 1,2,3,4,5,6,7,8. and since not all of the 3 structure are cyclic groups is it correct to use 'generator' to indicating the first element of 3 structures ?

2 way perfect shuffle will map $i \rightarrow 2i( mod (2n+1))$.

when $2n+1=3^k$, it's easy to derive the $k$ cyclic structures within 2 way perfect shuffle permutation. but the conclusion is based on this:

Thorem1 When p is odd prime and $g$ is the primitive root of $p^2$, we also have $\Rightarrow$ $g$ is also the primitive root of $p^k$

Def1 Euler function,$\phi(m)$ is defined as: for positive integers less than $m$,the number count which are co-prime with $m$

Def2 ord_m function,is defined as $Ord_m(a)=d, st: a^d\equiv1(mod m)$ minimum positive integer d

Def3 if $\phi(m)=Ord_m(a)$ then we call $a$ as primitive root of $m$

Property if $(a,m)=1$ and $a$ is the primitive root of $m$,Then $a$ is also the generator for cyclic group $(\Bbb Z/n\Bbb Z)^*$

for example when n=4,$2n+1=3^2$, ${1,2,3,4,5,6,7,8,9}$ contian 3 cyclic structures {1,2,4,5,7,8} {3,6} and {9}

the positive integer colletection coprime with 9 is {1,2,4,5,7,8},so we have $\phi(m)=6$, while $2^d\equiv1 (mod9)$ the minimum $d=6$ so 2 is primitive root of modulo 9,using the property of primitive root we can derive that 2 is the generator of cyclic group ${1,2,4,5,7,8}$

considering other strucutres that contain element of form $3^k$, these element must be of form $3^s2^t$ //questions1: why ?

when s=0, the structure is {1,2,4,5,7,8}

when s=1, the structure is {3,6}

when s=2, the structure is {9}

question2: can we call 1,3,9 generators of these cyclic structures as not all of them are cyclic groups, and how to better explain the cyclic structure whithin the perfect shuffle.

zinking
  • 1,089
  • 11
    Please write it in English. – Calmarius Aug 27 '13 at 07:28
  • Did you mean : http://math.stackexchange.com/questions/332760/prove-that-a-primitive-root-of-p2-is-also-a-primitive-root-of-pn-for-n1 – lab bhattacharjee Aug 27 '13 at 07:30
  • 13
    Лучше пишите по-русски. :-) – Boris Novikov Aug 27 '13 at 07:36
  • 2
    I can read chinese. The part about 定理1 is the theorem mentioned by lab bhattacharjee. However, there isn't any question in above sentences at all! – achille hui Aug 27 '13 at 07:39
  • come on! I just posted this a little early, recving so many vote downs. – zinking Aug 27 '13 at 08:09
  • @BorisNovikov: I can't read your comment, it is not fair. ;-) – Mikasa Aug 27 '13 at 11:18
  • @ Babak S. Oh, it's a joke. I wrote especially for OP: "It is better to write in Russian". – Boris Novikov Aug 27 '13 at 11:26
  • Is the question largely about the cycle structure of the permutation of the perfect shuffle? Depending on which half of the deck the top card comes from the answer is then either $(1)(8)(235)(476)$ or $(124875)(36)$. I don't get what the exactly 3 cyclic structures would refer to in either case. – Jyrki Lahtonen Aug 27 '13 at 11:49
  • But you should add some context, your own thoughts, etc. May be you did, as I can't read Chinese :-) Adding a translation-request tag. – Jyrki Lahtonen Aug 27 '13 at 11:50
  • @JyrkiLahtonen I have translated all comments, can you help me with it now ? thanks for making constructive comments. – zinking Aug 27 '13 at 13:05
  • 1
    If $2$ is a generator for $\mathbb{Z}{3^k}^*$, it will be a generator for $\mathbb{Z}{3^\ell}^*$ for all $\ell<k$ as well. So there will be one long cycle containing the numbers coprime to $3$, a slightly shorter one of those fivisible by $3$ but not divisible by $9$ et cetera. IOW the numbers $a$ with $\gcd(3^k,a)=3^\ell$ for a fixed $\ell, 0\le\ell\le k$, will each form a cycle. Altogether $k+1$ cycles. – Jyrki Lahtonen Aug 27 '13 at 14:45
  • You translated the question wrongly. There are supposed to be $k$ cycles, as seen in the second line where you replaced $k$ with $3$. – Calvin Lin Aug 28 '13 at 00:25
  • @JyrkiLahtonen this is key part I am asking, can you elaborate and make it an answer so that I can accept ? – zinking Aug 28 '13 at 01:53
  • Thanks for the vote of confidence. But that is the easy part. The hard part was to prove primitivity, and that is done in Calvin's +1 answer. To make sure that you get his last point: If we worked modulo $3^3=27$, that is $n=13$, then there would be one cycle of length 18 consisting of the numbers not divisible by three. There would also be a cycle of length six $$3\cdot1\to3\cdot2\to3\cdot4\to3\cdot8\to3\cdot7\to3\cdot5\to3\cdot1,$$ where the we have the numbers of the 6-cycle from the $n=4$ case multiplied by three. Similarly with numbers divisible nine. – Jyrki Lahtonen Aug 28 '13 at 07:40

1 Answers1

4

I'm assuming you understand the theory.

The 'hardest part' of proving this, is to show that $2$ is a primitive root of $\mathbb{Z}_{3^k} ^*$.

Since $\phi(3^k) = 2 \times 3^{k-1}$, this follows if we can show that $2 ^{3^{k-1} } \not\equiv 1 \pmod{3^k} $ and $2^{2 \times 3^{k-1} } \not \equiv 1 \pmod{3^k}$.

The first statement is easy. Expand $2$ as $3-1$, we get

$$ 2^{3^{k-1}} \equiv (3-1) ^ { 3^{k-1} } \equiv -1 \pmod{3^k} $$

The second statement is easy. Expand $2$ as $3-1$, we get

$$ 2^{2\times 3^{k-2}} \equiv (3-1) ^ { 3^{k-1} } \equiv -2\times 3^{k-1} + 1 \pmod{3^k} $$

Hence, $2$ is a primitive root. The cyclic nature of the shuffles are as Jyrki indicated.


Claim: The numbers that have the form $3^l \times r$, where $ 3 \not \mid r $ would form a cycle.

Proof: This follows directly from $2$ being a primitive root of $3^{k-l}$. Moves consist of going from $a$ to $2a\pmod{3^k}$.
From $3^l \times r$, we can move to all other $3^l \times r^*$, where $3 \not \mid r^*$.
It is also clear that we cannot move to any $3^{l^*} \times r^*$ for $l^* \neq l$.

Calvin Lin
  • 68,864