P R O C E E D I N G S OF T H E

Cambridge Philosophical Society

VOL. 65 MAY 1969 PART a

Proc. Camb. Phil. Soc. (1969), 65, 579

PCPS 65-59
Printed in Great Britain

Unique factorization rings

B Y C. R. FLETCHER
University College of Wales, Aberystwyih

In memory of Professor W. B. Pennington

{Received 23 February 1968)

1. Introduction. The concept of a unique factorization domain (UFD) has been
denned for commutative (e.g. (4) page 21) and non-commutative (l) integral domains.
We take the theory a stage further here by denning a unique factorization ring (UFR),
where throughout, a ring is understood to mean a commutative ring with identity,
possibly containing proper zero-divisors.
A UFR is a ring in which each non-unit element has a unique (in a certain sense)
irreducible decomposition. The zero element is included since proper zero-divisors
exist in general. We could also consider irreducible decompositions of units if we took
the (natural?) step of allowing units to be irreducibles. The subsequent work would
still be valid, for the 'relevant part' (see prior to Proposition 3) of the decomposition
of a non-unit (unit) element would contain no unit (merely one unit). However, we
have taken the traditional path and do not allow unit irreducibles.
We first show that we have a true generalization of a UFD by proving that an
integral domain is a UFR if and only if it is a UFD. We then see that the direct sum of
a finite number of UFR's is a UFR, and in particular, the direct sum of a finite number
of UFD's is a UFR. Not all UFR's are of this second type however (e.g. the ring of
integers modulo qm, q prime and m > 1). Finally we prove that a principal ideal ring
is a UFR.
2. Definition of a UFR. All elements introduced are members of a ring R unless
otherwise stated.
DEFINITION 1. A refinement of a factorization r = a1...an is obtained by factorizing
one or more of the a^s (i = 1, ...,n).
DEFINITION 2. A non-unit element peR is said to be irreducible if and only if each
factorization of p has a refinement containing p. If R is a domain, the above definition is
equivalent to the standard definition, except that here the zero element is irreducible.
37 Camb. Philos. 65, 3

[Link] Published online by Cambridge University Press

 580 C. R. FLETCHER
DEFINITION 3. The V -class of an element reR, denoted by U(r), is given by
U(r) = {/?e R\Pyr = r for some yeR).
The following results are easily proved.
PROPOSITION 1. (i) Iffi,yeU(r) then fiyeU{r).
(ii) If/3eU(rs)andreU(s) then /3eU(s).
(iii) U(rs) => U(r).
(iv) U(ar) = U(r) for a non-zero-divisor a.
DEFINITION 4. A U-decomposition of an element reR is a factorization of r
r = fa...p'k)fa...pn),
such that (i) p\, pt are irreducible i = 1,..., k, j = 1,..., n.
(ii) tie Ufa... pn) (t=l,...,*).
(iii) prfUfa...pr..pn) j= 1 n,
where A denotes that the term is omitted.
The two brackets will always be written in the above order. If the first bracket is
empty we write r = ( ) (px... pn). A necessary and sufficient condition for a U-decom-
position of an element to exist is that it has an irreducible decomposition. This condi-
tion is clearly necessary.
PROPOSITION 2. If reR has an irreducible decomposition then r has a V-decomposition.
Proof. Suppose r = px...pn, wherepi is irreducible i = 1, ...,n. Then either r = ( )
(Pi • • • Pn) a n d f has a [/-decomposition, or pt e Uip-^...fit...pn) for some i. If the latter,
suppose pt has this property, and similarly choose p2 say, where pze U{p3 ...pn).
Perform this operation repeatedly, choosing Pi,p2, -..,Pk until no further pi has this
property. We claim that r = (p1... pk) (pk+1.. • pn) is a {/-decomposition. From the
oonstruction prf U(pk+1... p,... pn) j = k+l,...,n, and pt e U(pi+1... pn) i = 1, ...,k.
By repeated application of Proposition 1 (ii) the proof is completed.
DEFINITION 5. Two elements a, beR are said to be associate if and only if a divides b
•and b divides a in R.
This condition is not equivalent to 'a = ub for some unit ueR'. For example let
R = F[x,y, z]l(x — xyz), where F is a field. Then x and xy are associates in R, but
x =}= xyu for any unit ueR.
In a [/-decomposition r = (p[...p'k)(p-^...pn) we call the factorizationp1...pn the
relevant part, the element s = px...pn the relevant element, and the set {p'v .. .,p'k} the
non-relevant set.
PROPOSITION 3. The relevant element of a U-decomposition of reR is an associate of r.
Proof. Suppose r = {p'1...p'k)(p1...pn).
Then p\ e Ufa... pn) for i = l,...,k,
and from Proposition 1 (i) , , TT, .
* Pi-PkeUfa...pn).
Hence r and pt... pn are associate.

[Link] Published online by Cambridge University Press

 Unique factorization rings 581
It follows immediately that the relevant elements of two [/-decompositions of r eR
are associate.
DEFINITION 6. Two U-decompositionsofreR

are said to be associate if and only if (i) m = n, (ii) after a suitable change in the numbering
of thefactors of the second U-decomposition, the elements Pi and qi are associate i = 1,..., n.
DEFINITION 7. A ring R is called a unique factorization ring (UFR) if and only if
(UFR 1) every non-unit element of R has a U-decomposition, (UFR 2) any two U-decom-
positions of a non-unit element of R are associate.
An element in i? having a {/-decomposition and satisfying UFR 2 is said to have
unique factorization. From Proposition 2 UFR 1 is satisfied whenever every non-unit
element of R has an irreducible decomposition. We first show that in the domain case
the definitions of UFR and UFD are equivalent. We assume the definition of UFD
given in (4), page 21.
THEOREM 4. An integral domain R is a UFR if and only if R is a UFD.
Proof. Suppose first that R is a UFR and consider only non-zero, non-unit elements-
of R. Then each E/-class is the class of units, and each non-relevant set is empty.
Clearly R is a UFD. Conversely if R is a UFD then every non-zero, non-unit element
of R has unique factorization. Since U(0) = R and 0$ U(s) for any s 4= 0, a J/-decom-
position of the zero element is of the form (p[... p'k) (0). Hence 0 has unique factoriza-
tion and R is a UFR.
3. The direct sum of UFR's. We prove that the direct sum of two UFR's is a UFR.
Induction may be used to extend the result to include a finite number of rings.
THEOREM 5. / / R and S are UFR's, then R@8isa UFR.
Proof. We may easily see that the irreducibles of R © S are all elements of the form
(u, q) and (p, v), where u, v are units and p, q are irreducibles in R and S respectively.
Consider any non-unit element (r, s)eR © S, where both r and s are not units. We
shall suppose that neither is a unit, but the proof is similar if this is not the case. Then
r and s have ([/-decompositions
r = (p'1...p'k)(p1...pn) a n d s = (q[... q\) (qx... qm).
Therefore (r, s) has the irreducible decomposition

and UFR 1 is satisfied.

Now suppose (r, s) has the two [/-decompositions
((r i, s[)... ft, s[)) ((r lf sj... (rm, sj) = ((pi, tr'j) ... (p'k, <r'k)) ((Pl, a,)... (pn, c r j ) .
Then (r'i,sdeU((r1,s1)...(rm,sm)) (i=l,...,l)
and r'i e U(rx... rm), s\ e U(s1... sm).
Also {ri,si)$U((rvs1)...(ri,si)...(rm,sm)) (i = l,...,m),
and exactly one of
ri$U(r1..Ji...rm) and si$U(s1...§i...sm)
37-2

[Link] Published online by Cambridge University Press

 582 C. R. FLETCHER
occurs. Suppose that Tr. A . .
rieU{r1...?i...rm) for i=
and r,£l7(r 1 ...f J ...r m ) for j = g +
Then st$ U(s1...§i... sm), si is not a unit, and rt is a unit. Similarly s} is a unit. Put
11 = rx...rg and z> = sg+1 ...sm, then itrff41 is irreducible in R and s?« is irreducible in *S.
ti PT1PP

*•= W —»i)(wB+1...rm) and s = &...«{)(«!... sffv)

are [/-decompositions of r and s respectively. Similarly
r = (Pi — Pk)(JlPh+i — Pn) and 5
= (<r'i — (rk)(<r1...<rhv),
where fi = px...ph and y = crft+1... o~n are units, are ^/-decompositions of r and s
respectively. Therefore m — g = n — h and g = h, i.e. m = w. r} and /9y are associate for
j = g + 1,..., n and st and o^ are associate for i = 1,... ,.g, after a suitable change in the
numbering of the factors. Thus (r^s^) is an associate of {p{,<rt) for i = 1, ...,»andi2©jS
is a UFR.
COROLLARY. If Rt is a UFR for i = 1 n then Rx® ... © Rn is a UFR.
In particular a finite direct sum of UFD's is a UFR. The converse is not true,
i.e. not every UFR may be expressed as a direct sum of UFD's. For consider the
ring of integers modulo qm, q prime and m > 1. This is principal and hence is a
UFR (Theorem 7, Corollary), but is not a direct sum of domains (proved by counting
the number of elements in a direct sum and equating characteristics).
4. Principal ideal rings. It is well known that a principal ideal domain is a UFD.
In the non-commutative case one can also obtain a unique factorization theorem
{(2), p. 34). We now show that a principal ideal ring (PIR) is a UFR. This extends the
work of Serbin(3) in the particular instance of a commutative ring. From (4), p. 245
we have the following definition and result.
DEFINITION 8. A PIR is called special if it has only one prime ideal P 4= R and if P is
nilpotent, that is Pm = (0) for some integer m > 0.
THEOREM 6. Every PIR is a (finite) direct sum of principal ideal domains and of
special PIR's.
Hence from Theorem 5, Corollary, we need only show that a special PIR is a UFR.
THEOREM [Link] special PIR then Risa UFR.
Proof. UFR 1 may be verified by considering the principal ideals generated by
elements not having an irreducible decomposition. This class has a maximal member
Ra, where a is not irreducible. Therefore a = a1...an where a does not divide ax and
Ra <=• Rai for all i (strict inclusion). Every ai thus has an irreducible decomposition,
and so does a. Contradiction. Suppose now that 22 has the unique prime ideal P with
P m = (0). If m = 1 then R is afield,and in particular is a UFR. If m > 1 put P = Ra.
Then a is the only irreducible in R up to associativity, all elements not in Ra are units,
and every non-unit non-zero r e R has the form r = uak, where it is a unit and k < m is
unique (4). Therefore R is a UFR.
COROLLARY. A PIR is a UFR.

[Link] Published online by Cambridge University Press

 Unique factorization rings 583
The author would like to express his thanks to the referee for communicating the
example which follows Definition 5.

REFERENCES
(1) COHN, P. M. Noncommutative unique factorization domains. Trans. Amer. Math. Soc. 109
(1963), 313-331.
(2) JACOBSON, N. The theory of rings (Amer. Math. Soc. 1943).
(3) SBBBIN, H. Factorization in principal ideal rings. Duke Math. J. 4 (1938), 656-663.
(4) ZAKISKI, O. and SAMUEL, P. Commutative algebra, vol. 1 (Princeton 1958).

[Link] Published online by Cambridge University Press