2

I am self-studying algebra using MIT's publicly online materials. The text for the course is the second edition of Artin's Algebra, Proposition 2.4.3 of which is stated without proof:

Let $x$ be an element of finite order $n$ in a group, and let $k$ be an integer that is written as $k = nq + r$ where $q$ and $r$ are integers and $0 \leq r < n$. Then

  1. $x^k = x^r$.

  2. $x^k = 1$ if and only if $r = 0$.

  3. The order of $x^k$ is $\frac{n}{d}$, where $d$ is the greatest common divisor of $k$ and $n$.

1 and 2 are very easy, and I'm willing to bet that 3 is easy as well -- but I'm nonetheless stuck. It's obvious that $(x^k)^{\frac{n}{d}} = 1$; I run into difficulty, however, trying to show that there is no positive integer $c < \frac{n}{d}$ such that $x^{kc} = 1$.

We certainly know that if there is such a $c$ then it must divide $\frac{n}{d}$. Moreover, we know that $n = n'd$, $k = k'd$ for positive integers $n', k'$. What I'm not seeing how to use, evidently, is the fact that $d = \gcd(n,k)$. I'm sure the answer is straightforward, but I'm afraid I'm missing it.

Any help would be appreciated...

Xam
  • 6,119

6 Answers6

2

The order of $x^k$ is the smallest positive integer $r$ such that $x^{kr}=1$, i.e. the smallest positive integer such that $kr$ is a multiple of $n$.

As $kr$ is also a multiple of $k$, this means $kr$ is the least common multiple of $k$ and $n$. Now we have $$kr=\operatorname{lcm}(k,n)=\frac{kn}{d}=k\,\frac nd,\quad\text{whence}\quad r=\frac nd.$$

Bernard
  • 175,478
  • This amounts to the dual of the proof in my answer, i.e. with $,[k,n]:= {\rm lcm}(k,n),$ we have $$(\ x^{\large k})^{\large r}!=1 \iff n\mid kr \iff k,n\mid kr\iff [k,n]\mid kr\iff [k,n]/k\mid r\iff n/(k,n)\mid r $$ using $\ k,n = kn.\ \ $ – Bill Dubuque Jul 27 '17 at 22:17
  • I agree with you. What's interesting, from my point of view, is that the main argument can be formulated in ordinary language. – Bernard Jul 27 '17 at 22:21
  • I have no idea what you mean by "ordinary language". – Bill Dubuque Jul 27 '17 at 22:23
  • In English (or any other language) – I mean with as few formulae as possible. Same difference as proving theorems of geometry with pure reasoning or with computations on coordinates. – Bernard Jul 27 '17 at 22:30
  • I'm not sure I follow your analogy. What would be a proof of this using "computations on coordinates"? – Bill Dubuque Jul 27 '17 at 22:33
  • It's only a vague analogy. I prefer prefer reasoning to computations. A good example in arithmetic is the traditional proof of Gauß' lemma, which relies on Bézout's identity, whereas there exists a beautiful proof, due to Cauchy if I remember well, which only uses that $\mathbf N$ is well-ordered. – Bernard Jul 27 '17 at 22:40
  • Both your proof and the version I gave in the comment above essentially rely on the universal property of the lcm, i.e. $,k,n\mid m\iff [k,n]\mid m.,$ The only difference is that you take the min at the start, and my comment takes the min at the end. Most proofs do end up employing this gcd / lcm property, but often less efficiently (e.g. doing both directions separately). But I wouldn't say they are as far apart as in your analogy. – Bill Dubuque Jul 27 '17 at 23:01
  • Perhaps you mean we could give a proof analogous to Zermelo's proof of unique factorization, i.e. eliminate all conceptual structure (gcds,lcms) by inlining all lemmas, resulting in a proof whittled down to the guts of arithmetic (assembly language vs. high-level language), using only induction and/or the division algorithm. This would surely be much less conceptual and more computational. If that's what you mean by "reasoning vs. computation", then I do agree. – Bill Dubuque Jul 27 '17 at 23:01
1

If $c$ is the order of $x^k$, then $(x^k)^c=1$, that is $x^{kc}=1$. But the order of $x$ is $n$ and therefore $n\mid kc$, that is $n'd\mid k'dc$, and this is equivalent to $n'\mid k'c$. Since $(n',k')=1$, it follows that $n'\mid c$, that is $\frac nd\mid c$. On the other hand, $(x^k)^{\frac nd}=x^{\frac kdn}=1$. But $c$ is the order of $x^k$. So, $\frac nd=c$.

1

Everything takes place in the cyclic subgroup generated by $x$, so it's not restrictive to assume $G=\langle x\rangle$, a cyclic group of order $n$.

Consider the homomorphism $f\colon\mathbb{Z}\to G$, $f(z)=x^z$, which is surjective and has kernel $n\mathbb{Z}$.

Now consider $g\colon\mathbb{Z}\to\mathbb{Z}$ defined by $g(z)=kz$. Since $f\circ g(1)=f(k)=x^k$, we have that the image of $f\circ g$ is $\langle x^k\rangle$ which we need to find the order of (the order of an element is precisely the number of elements in the cyclic subgroup it generates).

By general theory, the order is $l$, where $l\mathbb{Z}=\ker(f\circ g)$.

Now $$ l\mathbb{Z}=\ker(f\circ g)=\{z\in\mathbb{Z}:g(z)\in\ker f\}= \{z\in\mathbb{Z}:kz\in n\mathbb{Z}\} $$ If $d=\gcd(n,k)$, set $n'=n/d$ and $k'=k/d$. Then $$ kn'=k'dn'=k'n\in\mathbb{Z} $$ so $n'\in l\mathbb{Z}$ and $l\mid n'$. Conversely, $kl=nt$ for some $t$, so we have $k'dl=n'dt$, and $k'l=n't$ so $n'\mid k'l$. Since $k'$ and $n'$ are coprime, we see that $n'\mid l$.

Therefore $l=n'$ as we wished to show.

egreg
  • 238,574
1

Remember. If $d=\gcd(k,n)$ then $\frac kd$ and $\frac nd$ are relatively prime.

Now together prop 1) and 2) prove that if $x^j = 1$ then $j$ is a multiple of $n$.

So if $(x^k)^m = x^{km} = 1$ then $km$ is a multiple of $n$ and $\frac kdm$ is a multiple of $\frac nd$. And as $\frac kd$ and $\frac nd$ are relatively prime $m$ is a multiple of $\frac nd$.

So that's that.

If $m < \frac nd$ then $m$ is not a multiple of $\frac nd$ and $(x^k)^m \ne 1$.

So order of $x^k \ge \frac nd$. And as $(x^k)^{\frac nd} = (x^{n})^{\frac kd} = 1$ we have order of $x^k \le \frac nd$.

So order of $x^k$ is $\frac nd$.

fleablood
  • 124,253
1

Another option would be to use Proposition 2.3.5 c): There are some integers $r$ and $s$ with $d = rn+sk$. Suppose $x^k$ has order $m$. Then $x^{md}=x^{m(rn+sk)}=(x^n)^{mr}(x^k)^{ms}=1$ shows that $n \mid md$, so $\frac nd \mid m$. With $(x^k)^{\frac nd} = 1$ you have $m = \frac nd$.

Heraklit
  • 387
0

$(x^{\large k})^{\large j}\!=1\!\!\! \overset{\large \ \,\color{#0a0}{(1)}}\iff\! n\mid kj \iff n\mid kj,nj\!\! \overset{\large \ \,\color{#c00}{(2)}}\iff n\mid (kj,nj)\!=\!(k,n)j\iff n/(k,n)\mid j$

$\color{#0a0}{(1)}\ $ holds by $\, n = {\rm ord}\, x\ $ so $\ x^i = 1\iff n\mid i$

$\color{#c00}{(2)}\ $ uses the GCD Universal Property $\ n\mid a,b\iff n\mid(a,b)\,$ and $ $ GCD Distributive Law

Bill Dubuque
  • 272,048