nLab GCD domain

Redirected from "GCD domains".
Contents

Context

Arithmetic

Algebra

Contents

Definition

Definition

Let RR be an integral domain and let a,bRa,b \in R. We call gcd of aa and bb an element gcd(a,b)Rgcd(a,b) \in R such that:

  1. gcd(a,b)gcd(a,b) divides both aa and bb,

  2. for every element rRr \in R, if rr divides both aa and bb, then rr divides gcd(a,b)gcd(a,b).

In other words, a gcd of aa and bb is a meet in the preordered set (R,)(R,\mid) i.e. a cartesian product of aa and bb in (R,)(R,\mid) seen as thin category.

It follows from the properties of cartesian products that a gcd is unique up to isomorphism in the category (R,)(R,\mid) and that every element isomorphic to a gcd of aa and bb is also a gcd of aa and bb.

That is, if a gcd of aa and bb exists, then the gcd’s of aa and bb are exactly the products of this first gcd by any invertible element.

Definition

A GCD domain is an integral domain RR such that every pair a,bRa,b \in R of elements admits a gcd.

Properties

Equivalence between GCD domain and LCM domain

Definition

Let RR be an integral domain and let a,bRa,b \in R. We call lcm of aa and bb an element lcm(a,b)Rlcm(a,b) \in R such that:

  1. both aa and bb divides lcm(a,b)lcm(a,b),

  2. for every element rRr \in R, if both aa and bb divide rr, then lcm(a,b)lcm(a,b) divides rr.

As for the gcd, an lcm of aa and bb is just a join in the preordered set (R,)(R,\mid). Therefore, if an lcm of aa and bb exists, then the lcm’s of aa and bb are exactly the products of this first lcm by any invertible element.

The following proposition treats the interactions of gcd, lcm and 00.

Proposition

Let RR be an integral domain.

  • Let rRr \in R. Then gcd(r,0)=rgcd(r,0)=r and lcm(r,0)=0lcm(r,0)=0.

  • Let a,bRa,b \in R. Then, gcd(a,b)gcd(a,b) exists and is equal to 00 iff both aa and bb are zero.

  • Let a,bRa,b \in R. Then lcm(a,b)lcm(a,b) exists and is equal to 00 iff one of aa or bb is zero.

Proof

  • We have that rr divides both rr and 00. If ss divides both rr and 00, then in particular ss divides rr. Therefore gcd(r,0)=rgcd(r,0)=r. We have that both rr and 00 divides 00. If both rr and 00 divides ss, then in particular 00 divides ss. Therefore lcm(r,0)=0lcm(r,0)=0.

  • If gcd(a,b)=0gcd(a,b)=0, then both aa and bb are zero since gcd(a,b)gcd(a,b) divides both of them. If both aa and bb are zero, we already know from the previous point that gcd(a,b)=0gcd(a,b)=0.

  • If lcm(a,b)=0lcm(a,b)=0, then ab=0ab=0 since lcm(a,b)lcm(a,b) divides abab. It follows that either aa or bb is zero. Now, suppose that one of aa or bb is zero. We already know that both aa and bb divide 00. Suppose that cRc \in R is such that both aa and bb divides cc. One of a,ba,b being zero, it follows that 00 divides cc. We conclude that lcm(a,b)=0lcm(a,b)=0.

We will now proceed to prove that a GCD domain can equivalently be defined as an integral domain such that every pair of elements admits an lcm. First, we need a lemma about the compatibility of gcd’s with multiplication.

Lemma

Let RR be an integral domain. Let a,bRa,b\in R and let mRm \in R. Suppose that gcd(a,b)gcd(a,b) and gcd(ma,mb)gcd(ma,mb) exist. Then gcd(ma,mb)=mgcd(a,b)gcd(ma,mb)=m\cdot gcd(a,b) (up to multiplication by an invertible element).

Proof

If m=0m=0 then, we have gcd(ma,mb)=0=mgcd(a,b)gcd(ma,mb)=0=m\cdot gcd(a,b). We can therefore suppose that mm is nonzero.

  1. Since gcd(a,b)gcd(a,b) divides both aa and bb, we obtain that mgcd(a,b)m\cdot gcd(a,b) divides both mama and mbmb. Hence, mgcd(a,b)gcd(ma,mb)m\cdot gcd(a,b) \mid gcd(ma,mb).

  2. We know that mm divides both mama and mbmb. Therefore, mgcd(ma,mb)m \mid gcd(ma,mb). It follows that gcd(ma,mb)=mugcd(ma,mb)=mu for some uRu \in R. It follows that mumu divides both mama and mbmb i.e. there exists s,tRs,t \in R such that ma=msuma=msu and mb=mtumb=mtu. Since mm is nonzero, we deduce that a=sua=su and b=tub=tu. Thus uu divides both aa and bb. Hence, ugcd(a,b)u \mid gcd(a,b). By multiplying by mm, we obtain that gcd(ma,mb)mgcd(a,b)gcd(ma,mb) \mid m\cdot gcd(a,b).

From the two previous points, we deduce that gcd(ma,mb)gcd(ma,mb) and mgcd(a,b)m \cdot gcd(a,b) are associated.

We can now prove that:

Theorem

Let RR be an integral domain and let a,bR\{0}a,b \in R \backslash \{0\}. Then lcm(a,b)lcm(a,b) exists iff gcd(ma,mb)gcd(ma,mb) exists for every mR\{0}m \in R \backslash \{0\}.

Proof

\Leftarrow: Suppose that a,bR\{0}a,b \in R \backslash \{0\} are such that gcd(ma,mb)gcd(ma,mb) exists for every mR\{0}m \in R \backslash \{0\}. Then, in particular gcd(a,b)gcd(a,b) exists. We can thus write:

a=gcd(a,b)s a=gcd(a,b)s

and

b=gcd(a,b)t b=gcd(a,b)t

where s,tRs,t \in R.

We claim that gcd(a,b)stgcd(a,b)st is a lcm of aa and bb.

Indeed, both aa and bb divide gcd(a,b)stgcd(a,b)st. Moreover, suppose that qRq \in R is such that aqa \mid q and bqb \mid q. Then abqbab \mid qb and abqaab \mid qa. It follows that abgcd(qa,qb)=qgcd(a,b)ab \mid gcd(qa,qb)=q\cdot gcd(a,b). Thus gcd(a,b)gcd(a,b)stqgcd(a,b)gcd(a,b)gcd(a,b)st \mid q\cdot gcd(a,b). Since a,ba,b are not both zero, we have gcd(a,b)0gcd(a,b) \neq 0. We deduce that gcd(a,b)stqgcd(a,b)st \mid q. We have verified the two conditions for gcd(a,b)stgcd(a,b)st to be a lcm of aa and bb.

\Rightarrow: Suppose that a,bR\{0}a,b \in R \backslash \{0\} are such that lcm(a,b)lcm(a,b) exists. We can then write:

ab=lcm(a,b)s ab=lcm(a,b)s

and

lcm(a,b)=at lcm(a,b)=at

and

lcm(a,b)=bu lcm(a,b)=bu

where s,t,uRs,t,u \in R.

We claim that ss is a gcd of aa and bb.

We can thus write:

ab=ats ab=ats

and

ab=bus ab=bus

from which it follows that b=tsb=ts and a=usa=us. Therefore, ss divides both aa and bb.

Suppose now that vRv \in R is such that vv divides both aa and bb. Then vv is nonzero. Moreover, vabv \mid ab. Write ab=kvab=kv where kRk \in R. We have that aka \mid k (since vbv \mid b) and bkb \mid k (since vav \mid a). Therefore, lcm(a,b)klcm(a,b) \mid k. Write k=lcm(a,b)lk=lcm(a,b)l where lRl \in R. We obtain ab=lcm(a,b)lvab=lcm(a,b)lv. Hence, lcm(a,b)s=lcm(a,b)lvlcm(a,b)s=lcm(a,b)lv. Since both aa and bb are nonzero, we have that lcm(a,b)lcm(a,b) is nonzero. It follows that s=lvs=lv. Therefore vsv \mid s.

We have verified the two conditions for ss to be a gcd of aa and bb. We can thus write s=gcd(a,b)s=gcd(a,b).

Let mR\{0}m \in R \backslash \{0\}. We want to show that mgcd(a,b)m\cdot gcd(a,b) is a gcd of mama and mbmb.

We know that gcd(a,b)gcd(a,b) divides both aa and bb. It follows that mgcd(a,b)m \cdot gcd(a,b) divides both mama and mbmb.

Let vRv \in R be such that vv divides both mama and mbmb. Then vv is nonzero. Write ma=vαma=v\alpha from which we have mab=vαbmab=v\alpha b. We obtain that aαba \mid \alpha b (since vmbv \mid mb). Moreover, bαbb \mid \alpha b. It follows that lcm(a,b)αblcm(a,b) \mid \alpha b. By multiplying by vv, we get that lcm(a,b)vαbv=mablcm(a,b)v \mid \alpha bv=mab. Since ab=lcm(a,b)gcd(a,b)ab=lcm(a,b)gcd(a,b), we obtain that vmgcd(a,b)v \mid m\cdot gcd(a,b).

We have verified the two conditions for mgcd(a,b)m\cdot gcd(a,b) to be a gcd of mama and mbmb.

We can therefore state:
Corollary

Let RR be an integral domain. Then RR is a GCD domain iff every pair of elements of RR admits an lcm.

Properties of gcd’s and lcm’s

Let RR be a GCD domain. We already know that gcd(ma,mb)=mgcd(a,b)gcd(ma,mb)=m \cdot gcd(a,b) for every a,b,mRa,b,m \in R. The same is true for lcm’s.

Proposition

Let RR be a GCD domain. Let a,bRa,b \in R and let mRm \in R. Then lcm(ma,mb)=mlcm(a,b)lcm(ma,mb)=m\cdot lcm(a,b) (up to multiplication by a unit).

Proof

If m=0m=0 then the two sides of the equation are zero. We can thus suppose that mm is nonzero.

  1. We know that alcm(a,b)a \mid lcm(a,b), that blcm(a,b)b \mid lcm(a,b). Therefore, mamlcm(a,b)ma \mid m\cdot lcm(a,b) and mbmlcm(a,b)mb \mid m\cdot lcm(a,b). Hence lcm(ma,mb)mlcm(a,b)\lcm(ma,mb) \mid m\cdot lcm(a,b).

  2. We know that mmam \mid ma, therefore mlcm(ma,mb)m \mid lcm(ma,mb). It follows that we can write:

    lcm(ma,mb)=mu lcm(ma,mb)=mu

    where uRu \in R. Hence both mama and mbmb divides mumu. Since mm is nonzero, we obtain that both aa and bb divides uu. Hence, lcm(a,b)ulcm(a,b) \mid u. By multiplying by mm, we obtain that mlcm(a,b)lcm(ma,mb)m \cdot lcm(a,b) \mid lcm(ma,mb).

From the two previous points, we conclude that lcm(ma,mb)lcm(ma,mb) and mlcm(a,b)m \cdot lcm(a,b) are associated.

The compatibility of the product with gcd’s and lcm’s implies that the multiplicative monoid of an integral domain is a prelattice-ordered commutative monoid.

Other properties

Every GCD domain of dimension at most 1 is a Bézout domain.

Examples

See also

References

Last revised on August 19, 2024 at 15:06:27. See the history of this page for a list of all contributions to it.