Some Results On Completability in Commutative Rings

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

PACIFIC JOURNAL OF MATHEMATICS

Vol. 37, No. 2, 1971

SOME RESULTS ON COMPLETABILITY


IN COMMUTATIVE RINGS
MARION MOORE AND ARTHUR STEGER

In this paper, R always denotes a commutative ring with


identity. The ideal of nilpotents and the Jacobson radical of
the ring R are denoted by N(R) and J(R), respectively. The
vector [aίf •• ,αw] is called a primitive row vector provided
1 e (au ' ", an)', a primitive row vector [aίf , an] is called
completable provided there exists an n X n imimodular matrix
over R with first row αu , αn. A ring R is called a Z?-ring
if given a primitive row vector [αlf , αn], n ^ 3, and
(αlt '-,αn-2) %
there exists be R such that 1 e (αi, , αn-2, αn-\ + bαn). Simi-
larly, R is denned to be a Strongly B-ring (SB-ring), if d e
(αi, , αn)f n Ξ> 3, and (αi, , αw_2) 2 «/(•#) implies that there
exists be R such that de (αu , αn-2f αn-i + 6αw).
In this paper it is proved that every primitive vector over
a I?-ring is completable. It is shown that the following are
5-rings: ^-regular rings, quasi-semi-local rings, Noetherian
rings in which every (proper) prime ideal is maximal, and
adequate rings. In addition it is proved that R[X] is a B-ring
if and only if R is a completely primary ring. It is then
shown that the following are SB-rings: quasi-local rings, any
ring which is both an Hermite ring and a £>-ring, and Dedekind
domains. Finally, it is shown that R[X] is an SB-ring if and
only if R is a field.

LEMMA 2.1. Let R be α ring with A ϋ J(K), A an ideal of R.


Then R is a B-ring if and only if R/A is a B-ring.

Proof. Necessity: Let R be a B-ήng and let

(1 + A) e (a, + A, . •, an + A), n ^ 3
and
(αx + A, , α%_2 + A) g J(R/A) = J(R)/A ,

where α< e R, i = 1, n. Then 1 + A = Σ5U α ^ ί + A, 6{ G i2; hence


[αx, •• 5 α % ] is primitive. Since (au , an_2) g J(J?), it follows that
[αL + A, , α%_2 + A, (αn_! + 6αn) + A] is primitive for some be R.
Therefore, R/A is a 5-ring.
Sufficiency: Suppose i2/A is a 5-ring and suppose [α1? , αw] is a
453
454 MARION MOORE AND ARTHUR STEGER

primitive vector with (a19 , α%_2) §§ J(R). Hence [aλ + A, , an + A]


is a primitive vector; and, since (aί9 •• ,α n _ 2 ) ξ£J(R), we have (αx +
A, •••, α«_2 + -A) S J(R/A). Since S/A is a j?-ring, there exists δ +
A G R/A such that [αx + A, -, an^2 + A, (αn_! + δαj 4- A] is primitive.
It follows that (1 - u) e A £ J(i2), where
w—2

^ — Σ a{bi + (α w _! + δ α j δ % _ π δ^ G i?, i — 1, ,n — 1 .

Therefore, u is a unit of R; i.e., [α n ••-, α w _ 2 , an^ + δ α j is primitive.


THEOREM 2.1. If R is a B-ring then every primitive row vector
over R is completable.

Proof. Let R be a U-ring and let 1 e (α15 « , α j . The theorem


clearly holds for n = 1. If w = 2, then 1 = α ^ + α2τ/, x,yeR and the
matrix ί _ ^ ^M is unimodular; hence the result holds for n — 2.
Let n ^ 3, and suppose the result is established for k < n.

Case 1. If (a19 , αw_2) £ J(R) and 1 = Σ?=i α i δ ^ δ^ e ^^ th


^n 1 -
Σji=? ttίbi — an^ιbn_ι + α%δπ is a unit ueR. Let

(X2 U/ W _ 2

-6. &»_! 0 0 ... 0

0 Γ
Then F h a s determinant u, and it follows that [α n , an] is completable.

Case 2. If (ax, , a%__2) ^ ^(-B), then 1 e (alf , aw_25 aΛ_! + δa%),


for some be R. By the induction hypothesis, [a19 , a%_2, aΛ_i + δaπ]
is completable to an (n — 1) x (n — 1) unimodular matrix D. Let

0 0\
1 o I and let J3 = I D

\0 . . . 0 1,
Then BU is an n x n unimodular matrix whose first row is [aλ, « , an].

For convenience, we introduce the notation Z(A) to mean the set


of maximal ideals containing the ideal A; Z(a) will denote the set of
maximal ideals containing the element α.

THEOREM 2.2 If R is a ring such that for every ideal A §£ J(R),


Z(A) is finite, then R is a B-ring.
SOME RESULTS ON COMPLETABILITY IN COMMUTATIVE RINGS 455

Proof. The essentials of the proof are due to Reiner [4]. Let
1 e (αx, , an), n ^ 3, and (alf , an_2) g£ J(R). By the hypothesis on
R, Z(A) is finite where A = (αx, , α n _ 2 ). Let Z(A) = {Mίy > , Mr],
and note that iΐ be R and αίl_1 + ban ξ. Mi9 i = 1, , r, then [αL, ,
αw_2, a%_! + δαj is primitive.
For any M,eZ(A) such that aneMiy we have αw_x + ban^.Mi1 for
all 6ei2; otherwise, α ^ e J I ί i , and (αlf * , α j £ .Mi which contradicts
the hypothesis that [αx, * , α j is primitive.
For those M{eZ(A) for which an£M{, we have (αw, M*) = (1).
Hence there exists an %{ such that anXι = an^ (mod Mi). For these
Mif we can find (by the Chinese Remainder Theorem) an element be R
such that b = 1 — Xi (mod Mt). It follows that an^ + bang Mi, ί =
1, , r. Hence [α15 απ_2, αTO_! + 6αft] is primitive.

It follows from this theorem that quasi-semi-local rings and


Noetherian rings in which every proper prime ideal is maximal (in
particular, Dedekind domains) are 5-rings.

LEMMA 2.2. Let R be an F-rίng (i.e., a ring in which every


finitely generated ideal is principal) which satisfies the condition
that if 1 e (aL, a2, α3) with αx £ J(R) then 1 e (alf a2 + ba3) for some be R.
Then R is a B-ring.

Proof. Let 1 e (aL, , an), n ^ 3, and let (a) = (aίf , αw_2) ^


J(R). Hence 1 e (a, an_lf an). By the hypothesis on R, 1 e (a, an^ + 6αn);
hence, R is a J5-ring.

THEOREM 2.3. If R is an F-ring which satisfies the condition


that for every a,ceR ivith aέJ(R), there is an reR such that
Z(r) = Z(a) — Z(c), then R is a B-ring.

Proof. The proof is essentially the same as the proof of Theorem


5 of [2]. Let 1 e (a, 6, c), at J(R). By the hypothesis on R there
exists re R such that Z(r) = Z(a) ~ Z(c). Hence (c, r) = (1), so there
exists qe R such that 1 e (r, b + qc). We claim (α, b + qe) = (1).
Otherwise, there exists a maximal ideal M of R such that (α, 6 + <?c) £
Λf. Hence Λf e Z(α) and Me Zφ + gc). Since 1 6 (r, 6 + qc) it follows
that M$Z(r), so MςZ(c). But we now have MeZφ), contrary to
(α, 6, c) = (1). Therefore (α, 6 + ςfc) = (1). Lemma 2.2 completes the
proof.

THEOREM 2.4. Every adequate ring is a B-ring.

Proof. In the proof of Theorem 5.3 of [3], Kaplansky shows that


456 MARION MOORE AND ARTHUR STEGER

if R is an adequate ring and if 1 e (α, 6, c), a Φ 0, then there exists


qeR such that l e (α, 6 + qc). Since an adequate ring is an F-ring,
the result follows from Lemma 2.2.

THEOREM 2.5. Every π-regular ring is a B-ring.

Proof. If R is a ττ~regular ring, and if ae R/N(R), then by


Lemma 2.2 of [5], a is an associate of e + β, e an idempotent and β
a nilpotent of the ττ~regular ring R/N(R). Since β = 0, α = we, w a
unit of R/N(R). Therefore, a2 = u2e and u~ιa2 = ue = α. Hence,
R/N(R) is a regular ring and therefore an adequate ring ([1, Th. 11]).
Theorem 2.4 and Lemma 2.1 complete the proof.

2.6. Let D be an integral domain, K its quotient field.


THEOREM
Let R — {(alf , ak, α, a, •): α^ e K, a e D}, where k is a nonnegative
integer (k may be different for distinct elements of R). The operations
in R are component-wise addition and multiplication. If R is a B-
ring then D is a B-domain.

We illustrate t h e proof. Suppose R is a B-ήng and let 1 e (a, 6, c),


α, by c e D, 1 = αα' + bbf + cc\ L e t α = (1, α, α, •), 6 = (0, 6, 6, •),
c = (0,c, •), a' = (1, α', α', •), &' - (0, 6',^', •), ί ; = 0, c\ c', •)•
Then ϊ = aa' + 66' + c c \ If α e J(R), then 1 - α = (0, 1 - α, 1 - α, •)
is a unit of R. Since this is false, agJ(R), hence l e (α, 6 + yc) for
some y £ R. Therefore ϊ = ad + (6 + ^c)β, where d, c, e e i2. L e t d =
( ^ , - - > , d p , d , d , . . . ) , e = ( e t , •••, e 9 , e , β , • • • ) , ^ = ( i / i , • • • ^ 2 / r > ^ 1 / , • • • )
and let λ = max (1, p, g, r). In the (λ + l)st entry of ad + (6 + ^c)e,
we have ad + (6 + 2/c)e; i.e., 1 e (α, 6 + ?/c). Hence, ΰ is a S-domain.

THEOREM 2.7. iί[X] is a B-ring if and only if R is a completely


primary ring.

Proof. Sufficiency: Let R be a completely primary ring. Since


R/N(R) is a field and since (R/N(R))[X] ~ R[X]/N(R)[X], it follows
from Theorem 2.2 that R[X]/N(R)[X] is a B-ήng. Since N(R)[X] =
iV(J?[X]), the result follows from Lemma 1.2.1.
Necessity: Assume that R is not completely primary and that
R[X] is a i?-ring. Let r be a nonunit, nonnilpotent element of R.
Then 1 e (r, 1 + X, X2) and r £ J(i2[X]). By the assumption that R[X]
is a .B-ring, we have 1 e (r, 1 + X + X2/(X)) for some /(X)ei2[X].
Let α denote the image of a e R under the natural homomorphism of
R[X] onto (R/rR)[X]. Then ϊ e (0, ϊ + X + X2f(x)) and Ϊ + X +
X2/(X) is a unit of (i?/ri?)[XJ. This is a contradiction since the coe-
fficient of X is not nilpotent.
SOME RESULTS ON COMPLETABILITY IN COMMUTATIVE RINGS 457

Since R[X] cannot be completely primary, (clearly, X is neither


a unit nor a nilpotent) it follows that for every ring R, R[X, Y] =
R[X][Y] is not a 5-ring.

3* Strongly brings* We now turn our attention to the study


of SB-rings. Our main objective here is to compare the theory of
this particular subclass of 5-rings with that of 5-rings given in the
last section.

LEMMA 3.1. R is an SB-ring if and only if for every s, cu c2,


c3e R with s e (c19 c2, c3) and cγ £ J(R), it follows that s e (c15 c2 + bc3)
for some b e R.

Proof. The necessity clearly follows from the definition of an Sp-


ring.
Sufficiency: Let r e (a19 , an), n ^ 3, with (aίy , an_2) §= J(R).
Without loss of generality, we may assume that an_2 £ J(R). Suppose
an(
r = Σ * = 1 aiχi i let s = an_2xn_2 + an_Lxn^ + anxn. Then r e (aL, ,
α%_3, s) and s £ (an_2, an^, an). Since απ__2 e J(R), s e (an_2, a n _ x + ban) for
some be R. Therefore r e (a,, , α%_39 s) S (a,, , an^ + ban), and
the proof is complete.

In view of Lemma 3.1, we need only consider triples instead of


arbitrary π-tuples in our study of SB-rings.

LEMMA 3.2. The homomorphίc image of an SB-ring is an SB-


ring.

Proof. Let R be the image of R under the homomorphism φ, and


let d £ (α15 α2, α3) with aι e J(R), a19 a2, azi d e R. Suppose d = Σ L ^ ,
Xi e R and let a{φ — a{, Xiφ = x%, i = 1, 2, 3. Let d = Σί=i ^ i ^ Since
(J(R))Φ S J(R), we have a.gJiR); hence, de (a1? a2 + δα3) for some
be R. Since dφ — d, we have d e (aly a2 + δ"α3), where bφ = b. Hence
R is an SB-ring.

THEOREM 3.1. Every quasi-local ring is an SB-ring.

Proof. Let d e (αx, α2, α3), with αx G J(i2), iϊ a quasi-local ring. Since
aι $ J(R), aγ is a unit of R; hence, d e (α15 a, + 6α3) = (1) for every be R.

LEMMA3.3. Let A — (au , α j , ^ ^ 3, δβ an ideal in a Dedekind


domain R. If B = (αlf , αΛ_2) ^ (0), ίAs^ A = (αx, , αn_2, αw_! +
δαj /or some be R.
458 MARION MOORE AND ARTHUR STEGER

Proof. Let A = Π U Mp and let B = Π U Ml* be the representa-


tions of the ideals A and B as a product of powers of distinct maximal
ideals. Since B i i , we may order the Mt so that 0 ^ a{ < βt for
1 <^ i <; r, and α< = βi for r + 1 <Ξ ΐ ^ ί. Let I <L k ^ r. We claim
that either an_γ or α n does not belong to Mik+ι. For suppose both αn_j.
and αw belong to Mcίk'\ Then A £ Mϊk+\ a contradiction. Since the
Mjc^1 are relative prime, the Chinese Remainder Theorem guarantees
the existence of a b e iί satisfying:
6 = 0 (mod M?*+1) if αΛ_x £ M^1
b = 1 (mod Mΐ'- 1 ) if α ^ e M^+1 ,
for k = 1, 2, , r. It follows that αn_! + δα% e M?* +ι for A; = 1, 2, , r.
Let (α15 , α%_2, a,,,! + 6aΛ) = Π U Mϊ'€. Since (ax, , an_2, aΛ_! + 6aΛ) S
A = ΠLi ΛίΓS it follows that μ{ ^ ^ , i = 1, 2, . . . , ί. Since £ = Πi=i
Λf/< g Πi=i Miι C Πi=i ^? - ί = 4> and since ^ = aif r + 1 ^ i £ t, it
follows that μι = βz = aiy r + 1 ^ i S t. lί μ{> a{ for some i with
1 ^ i ^ r, then αn_! + ban e Ml'* C M?<+1, a contradiction. Hence, /^^ =
aiy ί = 1, 2, , t. Equivalents, (α1? , αw_2, α n ^ + 6αw) = A.
As an immediate consequence, we have:

THEOREM 3.2. A Dedekind domain is an SB-ring.

LEMMA 3.4. Let Rbe a B-ring, let e = e2 e R, and let e e (aly , an)
with (alf , αw_2) g J(R), n ^ 3. ΓAe7^ e e (α1? , α w _ 2 , an^ + 6αw) /o?*
some be R.

Proof. Since t h e case β — 1 is covered by t h e hypothesis, we


may assume eφl. L e t e = Σ ? = 1 a{x{ = Σ?=i (^β)(^β). Hence, 1 =
(αxβ + 1 - e)(^β + 1 - e) + Σ t 2 (αίβ)(α;ΐe). Thus,
1 e (ate + 1 — e, α2e, , αwe) .
If aγe + 1 — e e J(iϊ), then 1 — (αxβ + 1 — e) = e(l — aλ) is a unit of R,
a contradiction since e = β2, β ^ 1. Thus, since R is a i?-ring, we have
1 G (aλe -r 1 - e, α2e, , α%_2β, α n _!e + 6αTOβ) for some 6 e R. Therefore,
e e (a.e, a2e, , αw_2β, αw_1e + bane) £ (αx, α 2 , , α%_2, α n _! + δ α j .

COROLLARY. If R is a regular ring then R is an SB-ring.

Proof. The result is immediate from Theorem 2,5 and Lemma


3.4; since, for every r e R, r is an associate of some idempotent ee R
([1, Lemma 10]).

THEOREM 3.3. // a B-ring R is also an Hermite ring, then R


is an SB-ring.
SOME RESULTS ON COMPLETABILITY IN COMMUTATIVE RINGS 459

Proof. L e t de (au a2f α3) = (α), aL£ J{R). By Corollary 5 of [1],


t h e r e exist b19 b2, b3 such t h a t aι = bLa, α 2 = δ 2 α, α 3 = δ 3 α, and (b19 b2, δ3) —
(1). Since ϋ? is a 5-ring and since b^J(R), t h e r e exists a qeR such
t h a t (δ1? δ2 + qb3) = (1). Therefore, (α) = ( δ ^ , δ2α + gδ3α) = (αL, α 2 + qα3).
Hence, de(αιy α2 Λ- qα3).

COROLLARY. Every adequate domain is an SB-ring.

Proof. An adequate domain is both an F-domain and a B-ring.


Since every F-domain is an Hermite ring, the result follows from
Theorem 3.3.

COROLLARY. // R is an F-ring with infinitely many maximal


ideals and, if for every ideal A ξ£ J(R), Z(A) is finite, then R is an
SB-ring.

Proof. R is necessarily a S-ring by Theorem 2.2. By the proof


of Corollary 2 of [2], R is also an Hermite ring. Theorem 3.3 com-
pletes the proof.

THEOREM 3.4. R[X] is an SB-ring if and only if R is a field.

Proof. The sufficiency follows from Theorem 3.2. To prove the


necessity, let r eR,r Φ 0. Then r e (X2, X, r) and X2 e J(R[X\). If
R[X] is an SB-ring then re(X\X+rb(X)) for some b(X) e R[X].
2
Let r - X f[X) + (X + rb(X))g(X), where f(X) and g(X) e R[X], and
let fi, git bi represent the coefficient of X* in the polynomials f(X),
g(X), b(X), respectively. Equating coefficients in the above equation
gives r = rbQgQ and 0 = gQ + rφ.g, + g,bv). Hence r divides gQ and
therefore r — r2k for some ke R. Hence rk — (rk)2; therefore, rk is
an idempotent of R. Since R[X] is a B-ring, R must be a completely
primary ring by Theorem 2.7. It follows that the idempotent rk is
2
either 0 or 1. Since rk = 0 and r = r k imply r = 0, we conclude
that rk = 1; i.e., r is a unit of R. Hence, R is a field and the proof
is complete.

REFERENCES

1. L. Gillman and M. Henriksen, Some remarks about elementary divisor rings, Trans.
Amer. Math. Soc. 82 (1956), 362-365.
2. M. Henriksen, Some remarks on elementary divisor rings II, Michigan Math. J. 3
(1955-56), 159-163.
3. I. Kaplansky, Elementary divisors and modules, Trans. Amer. Math. Soc. 66 (1959),
464-491.
4. I. Reiner, Unimodular complements, Amer. Math. Monthly 6 3 (1956), 246-247.
460 MARION MOORE AND ARTHUR STEGER

5. A. Steger, Elementary factorization in π-regular rings, Canad. J. Math. 18 (1966),


307-313.

Received October 30, 1969.

UNIVERSITY OF TEXAS AT ARLINGTON


UNIVERSITY OF NEW MEXICO

You might also like