3

Do you know of a proof of $[a,b](a,b)=ab$ in $\mathbb Z$ that doesn't use prime factorization?

To be more precise let's strip all unnecessary properties and leave only the bare bones of divisibility: on a commutative monoid $M$ with the cancellation law and for which the g.c.d and l.c.m of two elements exists, does $[a,b](a,b)\sim ab$ still hold?

($\sim$ means equality up to a unit factor, also notice that this allows the existence of an infinite chain of divisors, otherwise it's known that M is factorial, see for example theorem 2.22 in Jacobson Basic Algebra I)

Edit: To clarify the notation, $[a,b]$ is the lcm of $a$ and $b$, while $(a,b)$ is the gcd.

Fra
  • 899

2 Answers2

2

We'll use $x,y$ for the $a,b$ in the post, and retain the notation $(x,y)$ for the gcd and $[x,y]$ for the lcm, which does appear in a few number theory texts. Put $(x,y)=d,\ [x,y]=m$ and write $x=da,\ y=db.$ Then $(a,b)=1$ follows (meaning it is a unit, gcd and lcm only being defined up to units). For if $k|a,\ k|b$ then also $kd|x,\ kd|y$ and then $kd|(x,y)=d$ giving $k|1$ i.e. $k$ is a unit.

We'll need a Lemma for later:

If $(a,b)=1$ then $[a,b]=ab.$

Let $e=[a,b]$, so that since $ab$ is one of the common multiples of $a,b$ we have $e|ab.$ So we may write $ek=ab.$ From $e=[a,b]$ there are $u,v$ for which $au=e,\ bv=e.$ Then $auk=ab$ gives $uk=b$ so $k|b,$ and similarly $bvk=ab$ gives $vk=a$ so $k|a.$ Then from $k|a,b$ follows $k|(a,b)=1,$ i.e. $k$ is a unit, and $[a,b]=e=ab$ up to the unit $k.$

Now define $m=[x,y]=[ad,bd]$ and consider the term $abd.$ It is a common multiple of $ad$ and $bd$ and so $m|abd.$ On the other hand, $m$ is a multiple of $d$ since each of $ad,bd$ divide $m,$ so we may write $m=dz.$ Then from $ad|m=dz$ follows $a|z,$ and from $bd|m=dz$ follows $b|z.$ So each of $a,b$ divides $z,$ hence so does $[a,b].$ We are here in the situation that $(a,b)=1,$ so by the Lemma we have $[a,b]=ab,$ and so $ab|z,$ and there is $t$ with $abt=z.$ But then $(abd)t=dz=m,$ so that $abd|m.$

We now have both $m|abd$ and $abd|m$ making these associates. Finally from that we get $md=(ad)(bd)=xy,$ up to a unit, i.e. $[x,y]\cdot(xy)=xy$ (again up to a unit).

coffeemath
  • 30,190
  • 2
  • 34
  • 53
2

Let $M$ be a commutative cancellative monoid in which gcds and lcms exist.

Lemma 0. If two elements $x,y$ of $M$ satisfy $x|t$ if and only if $y|t$, for all elements $t$, then $x$ and $y$ are associated.

Proof. For $t=x$ we see that $x|x$ implies $y|x$. By symmetry, we also get $x|y$. Hence, $x,y$ are associated. (Actually this is a special case of the Yoneda Lemma.)

Lemma 1. If $a,b,t$ are elements of $M$, then $\mathrm{gcd}(a \cdot t,b \cdot t)$ and $\mathrm{gcd}(a,b) \cdot t$ are associated.

Proof. Clearly $\mathrm{gcd}(a,b) \cdot t$ divides $a \cdot t$ and $b \cdot t$ and hence also $\mathrm{gcd}(a \cdot t,b \cdot t)$. Conversely, observe that $t$ divides $\mathrm{gcd}(a \cdot t,b \cdot t)$ since it divides $a \cdot t$ and $b \cdot t$. Write $\mathrm{gcd}(a \cdot t,b \cdot t) = c \cdot t$. Then $c$ divides $a$ and $b$, and hence $c$ divides $\mathrm{gcd}(a,b)$. So $\mathrm{gcd}(a \cdot t,b \cdot t)$ divides $\mathrm{gcd}(a,b) \cdot t$.

Proposition. If $a,b$ are elements of $M$, then $\mathrm{lcm}(a,b) \mathrm{gcd}(a,b)$ is associated to $ab$.

Proof. So let $t$ be any element of $M$. Then:

$\phantom{.--} \mathrm{lcm}(a,b)/b | t$

$\Longleftrightarrow$ $\mathrm{lcm}(a,b) | bt$

$\Longleftrightarrow$ $a | bt$ and $b | bt$

$\Longleftrightarrow$ $a | bt$

$\Longleftrightarrow$ $a | bt$ and $a|at$

$\Longleftrightarrow$ $a | \mathrm{gcd}(at,bt) \stackrel{\text{Lemma 1}}{=} \mathrm{gcd}(a,b)t$

$\Longleftrightarrow$ $a/\mathrm{gcd}(a,b) | t$

Hence, Lemma 0 implies that $\mathrm{lcm}(a,b)/b$ is associated to $a/\mathrm{gcd}(a,b)$.

  • Why the downvote? I've explained every detail, right? And the argument is actually very quick. – Martin Brandenburg Nov 25 '14 at 09:56
  • I agree, why the downvote? The proof look good to me (though the $/$ notation might be confusing). I will accept the other one mainly because it was the first one and because it's more constructive. But that it just a matter of personal taste. – Fra Nov 25 '14 at 16:40
  • This answer does use less manipulations than mine, +1. – coffeemath Nov 25 '14 at 21:03
  • The proof seems fine - don't know why it was downvoted. – goblin GONE Nov 20 '16 at 11:37
  • Both of the standard results were already proved here many times, many years ago, e.g. here and here, which are easily found by obvious searches. Perhaps that explains the downvote (not by me). $\ \ $ – Bill Dubuque Sep 15 '24 at 17:00