Proof check (I swear, if I ever misclick again...)

Boi

New member
Joined
Feb 14, 2023
Messages
30
View attachment 39388
I'll be using o(g)o(g) to denote the order of an element gg, ee to denote the neutral element and g\langle g \rangle to denote the cyclic group of gg.
For the proof I'll have to prove 2 lemmas first (they're very obvious results, but they are not stated in the book, so I've felt the need to prove them).
Lemma 1: If GG is a group, gGg \in G, o(g)=no(g) = n, p,qZp, q \in \mathbb{Z} and pq  (mod  n)p \equiv q \; (\mathrm{mod} \; n), then gp=gqg^{p}=g^{q}.
Proof of Lemma 1: if pq  (mod  n)p \equiv q \; (\mathrm{mod} \; n), then, by defenition of congruence modulo n, pq=knp-q=kn, where kk is an integer. It follows that p=q+knp = q+kn. So, gp=gq+kn=gqgkn=gq(gn)k=gqek=gqg^{p}=g^{q+kn}=g^{q} \cdot g^{kn}=g^{q} \cdot (g^{n})^{k}=g^{q}\cdot e^{k}=g^{q}.
Lemma 2: If n,mZn, m \in \mathbb{Z}, d=gcd(n,m)d=\mathrm{gcd}(n, m), n0=ndn_0=\frac{n}{d} and m0=mdm_0=\frac{m}{d}, then n0n_0 and m0m_0 are coprime.
Proof of Lemma 2: suppose the opposite is true: n0n_0 and m0m_0 are not coprime and they have a common divisor f>1f>1. So, n0=fsn_0=fs and m0=frm_0=fr (ss and rr are integers). By "unwrapping" the definitions of n0n_0 and m0m_0 and multiplying by dd on both sides we get n=dfsn = dfs and m=dfrm=dfr. Now we have reached a contradiction: we assumed dd to be the greatest common divisor of nn and mm, but D=dfD=df clearly divides both nn and mm without dividing dd (since we assumed f>1f>1).
The proof itself: Without the loss of generality I am going to assume that mnm \ge n. I'll denote the element of order nn by aa and the element of order mm by bb. Since GG is abelian, for all jZj \in \mathbb{Z} (ab)j=ajbj(ab)^{j}=a^{j}b^{j}. By using Lemma 1 we can represent every element of ab\langle ab \rangle with a pair of integers (x,y)(x, y), 0xn10 \le x \le n-1, 0ym10 \le y \le m-1 and for which ther exists such an integer zz, that zx  (mod  n)z \equiv x \; (\mathrm{mod} \; n) zy  (mod  m)z \equiv y\; (\mathrm{mod} \; m). I want to replace the last requirement with the following: xx and yy must be congruent modulo dd (where d=gcd(n,m)d= \mathrm{gcd}(n, m)). Let me show that these are equivalent.
\Rightarrow
If zx  (mod  n)z \equiv x \; (\mathrm{mod} \; n) and zy  (mod  m)z \equiv y \; (\mathrm{mod} \; m), then z=x+nuz=x+nu and z=y+mvz=y+mv, so x+nu=y+mvx+nu=y+mv (u,vZu,v \in \mathbb{Z}. It follows that xy=mvnux-y=mv-nu. If we pull out dd from nn and mm we get xy=d(m0vn0u)=dhx-y=d(m_0v-n_0u)=dh. Hence xy  (mod  d)x \equiv y \; (\mathrm{mod} \;d).
\Leftarrow
If xy  (mod  d)x \equiv y \; (\mathrm{mod} \; d), then xy=dhx-y=dh, hZh \in \mathbb{Z}. Now, using Lemma 2 and Bezout's identity we get 1=n0u+m0v1=n_0u+m_0v. Substituting u=tu=-t and multiplying both sides by hh we get h=m0vn0th=m_0v-n_0t. So, xy=dh=d(m0vn0t)=mvntx+nt=y+mvx-y=dh=d(m_0v-n_0t)=mv-nt \Rightarrow x+nt=y+mv. Letting z=x+ntz=x+nt we get the exact zz we need.
With the change of the last requirement it's easier to count the elements of ab\langle ab \rangle. Let's first choose xx. Remembering our constraints (0xn10 \le x \le n-1) we have nn choices for xx. Now, to count choices for yy let me first the number of choices for yy when x=0x=0. If x=0x=0, then yy should take the form dkdk to meet the condition. There would be md\frac{m}{d} such choices (0,  d,  2d,  ...,(md1)d0, \;d, \; 2d, \; ...\: , (\frac{m}{d}-1)d). When 0<x0 < x, we can just shift all these numbers by xx... except for the larger ones. Well, to not change the total number of choices for yy we should make a new valid choice for every failed one. So, if m1<x+dsm-1 < x+ds, then d(m0s)1<x1<xd(m0s)d(m_0-s)-1 < x \Rightarrow -1 < x- d(m_0-s). Because xd(m0s)x-d(m-0-s) is an integer, we can change the last inequality to 0xd(m0s)0 \le x-d(m_0-s), making xd(m0s)x-d(m_0-s) a viable choice for yy. Finally, we have nn choices for xx, md\frac{m}{d} choices for y, giving us a total of nmd\frac{nm}{d} or lcm(n,m)\mathrm{lcm}(n, m) elements in ab\langle ab \rangle. ab\langle ab \rangle having lcm(n,m)\mathrm{lcm}(n, m) means that abab, an element of GG, has order lcm(n,m)\mathrm{lcm}(n, m). QED.
I hope it's comprehensible.
 
For the record. The statement is:

Let G G be an abelian group, and suppose G G has elements of order n n and m m; then there is an element of order the least common multiple of n n and m. m.

The proofs of the lemmas are correct, but you lost me in the proof of the statement when x,y x,y came into play. I think it is way too complicated. We have the equation gcd(n,m)lcm(n,m)=nm. \operatorname{gcd}(n,m)\cdot \operatorname{lcm}(n,m)=n\cdot m. Now, use this and your remark that (ab)j=ajbj (ab)^j=a^j b^j and calculate (ab)lcm(n,m). (ab)^{\operatorname{lcm}(n,m)}.
 
We have the equation gcd(n,m)lcm(n,m)=nm. \operatorname{gcd}(n,m)\cdot \operatorname{lcm}(n,m)=n\cdot m. Now, use this and your remark that (ab)j=ajbj (ab)^j=a^j b^j and calculate (ab)lcm(n,m). (ab)^{\operatorname{lcm}(n,m)}.
Okay, I'll try to do that. I'll come back once I succeed.
But also, even if the proof is overcomplicated, is it correct?
 
Okay, I'll try to do that. I'll come back once I succeed.
But also, even if the proof is overcomplicated, is it correct?

As I already said, you lost me with the introduction of x,y. x,y. What are they? How are they determined, and why is xy(modnm), x\equiv y\pmod{nm}, and last but not least, how do they "represent" ab ab ?
 
As I already said, you lost me with the introduction of x,y. x,y. What are they?
how do they "represent" ab ab ?
So, any element of ab\langle ab \rangle is abab raised to some integer power zz. Because GG is abelian, we have (ab)z=azbz(ab)^{z}=a^{z}b^{z}. Now, if we reduce zz modulo nn (for aa) and modulo mm (for bb) we would get axbya^{x}b^{y} where 0xn10 \le x \le n-1, 0ym10 \le y \le m-1, zx  (mod  n)z \equiv x \; (\mathrm{mod} \; n) and zy  (mod  m)z \equiv y \; (\mathrm{mod} \; m). This is where xx and yy come from and how they "represent" an element of ab\langle ab \rangle.
why is x≡y(modnm), x\equiv y\pmod{nm}, x≡y(modnm)
I don't know where you got that congruence from.
Btw, thanks for your suggested way of solving the task, it seems much cleaner, although I hadn't wrote the proof with it yet
 
Btw, thanks for your suggested way of solving the task, it seems much cleaner, although I hadn't wrote the proof with it yet
It is only one line with maximal three equality signs. Plus the equation nm=gcd(n,m)lcm(n,m) nm=\operatorname{gcd}(n,m)\operatorname{lcm}(n,m) of course. So if you want to prove something, you should prove that equation.
 
only one line with maximal three equality signs
well, maybe not tha clean.(o(a)=n,o(b)=m,d=gcd(n,m),n0=nd,m0=mdo(a)=n, \: o(b)=m, \: d=\mathrm{gcd}(n, m), \: n_0=\frac{n}{d}, \: m_0=\frac{m}{d}) I mean, sure (ab)lcm(n,m)=anm0bn0m=(an)m0(bm)n0=e(ab)^{\mathrm{lcm}(n, m)}=a^{nm_0}b^{n_0m}=(a^{n})^{m_0}(b^{m})^{n_0}=e But that does not meant that o(ab)=lcm(n,m)o(ab)=\mathrm{lcm}(n, m).
 
Some remarks on your proof.

  • It would have been clearer if you had used either j j or z z and explained it as in your last post. I still have trouble dealing with those many variables.
  • "Should" should not occur in a proof. Readers do not want to read what should be, only what is. That's why many proofs are scribbled one way but written down backward. If you say "should," then I either stop reading or scribble the possible cases myself. And that's painful, particularly with those many variables. I'm currently stuck at your first "should". It asks me "why should it be?" and "what may I assume as given and what as assumed?"
  • Greater than or less than shouldn't occur at all. We only need integers because it is about divisibility, and signs don't matter.
  • I think you did the same thing as in my line of equations, only more complicated.
 
well, maybe not tha clean.(o(a)=n,o(b)=m,d=gcd(n,m),n0=nd,m0=mdo(a)=n, \: o(b)=m, \: d=\mathrm{gcd}(n, m), \: n_0=\frac{n}{d}, \: m_0=\frac{m}{d}) I mean, sure (ab)lcm(n,m)=anm0bn0m=(an)m0(bm)n0=e(ab)^{\mathrm{lcm}(n, m)}=a^{nm_0}b^{n_0m}=(a^{n})^{m_0}(b^{m})^{n_0}=e But that does not meant that o(ab)=lcm(n,m)o(ab)=\mathrm{lcm}(n, m).
It goes as follows:
(ab)lcm(n,m)=alcm(n,m)blcm(n,m)=(an)mgcd(n,m)(bm)ngcd(n,m)=(e)mgcd(n,m)(e)ngcd(n,m)=e. (ab)^{\operatorname{lcm}(n,m)}=a^{\operatorname{lcm}(n,m)}b^{\operatorname{lcm}(n,m)}=\left(a^n\right)^{\frac{m}{\operatorname{gcd}(n,m)}}\left(b^m\right)^{\frac{n}{\operatorname{gcd}(n,m)}}=\left(e\right)^{\frac{m}{\operatorname{gcd}(n,m)}}\left(e\right)^{\frac{n}{\operatorname{gcd}(n,m)}}=e. which means o(ab)lcm(n,m). o(ab)\,|\,\operatorname{lcm}(n,m).

And, yes, you are right. We also need to show that lcm(n,m)o(ab). \operatorname{lcm}(n,m)\,|\,o(ab).
 
Here is my version:

If o(ab)lcm(n,m) o(ab)\,|\,\operatorname{lcm}(n,m) then
o(ab)=qlcm(n,m)=sn=tme=(ab)o(a,b)=ao(a,b)bo(a,b)=atm=bsno(a)=ntm=o(a,b)o(b)=msn=o(a,b)lcm(n,m)o(a,b)\begin{array}{lll} o(ab)&=q\cdot\operatorname{lcm}(n,m)=sn=tm\\[10pt] e&=(ab)^{o(a,b)}=a^{o(a,b)}b^{o(a,b)}=a^{tm}=b^{sn}\\[10pt] &\Longrightarrow o(a)=n\,|\,tm=o(a,b)\wedge o(b)=m\,|\,sn=o(a,b)\\[10pt] &\Longrightarrow \operatorname{lcm}(n,m)\,|\,o(a,b) \end{array}
I hope this is correct.
 
Top