ROCKY MOUNTAIN

JOURNAL OF MATHEMATICS
Volume 20, Number 1, Winter 1990

A NOTE ON VALUATIONS, p-PRIMES

AND THE HOLOMORPHY SUBRING
OF A COMMUTATIVE RING

K.G. VALENTE

Introduction. In this note we use a recent result from the theory of

valuations to identify p-primes in a special class of commutative rings.
Our identification process will also provide the associated field with
p-prime in each case. We refer the reader to [5], for all definitions
pertaining to the p-prime invariants which, due to their length, will
not be included here. At the conclusion of this paper we will apply our
new information on the structure of all 0-primes to obtain a result on
holomorphy subrings which is closely related to some earlier work by
the author (see [4]).
We take this opportunity to introduce some notation which we will
employ throughout this paper. Firstly, all rings will be understood to
be commutative and unitary. For an arbitrary subset, S, of a ring R,
set
I(R, S) = {r ∈ R|rS ⊆ S}
and
(R : S) = {r ∈ R|rR ⊆ S}.
We let U (R) denote the group of units of R, and, in the case R is a
domain, qf(R) will represent the field of fractions of R. For I a prime
ideal of R, let RI be the localization of R at I. If R is assumed to
be a local ring with maximal ideal M , then k(R) and π will be used,
respectively, to signify the residue class field R/M and the natural
projection from R onto this field. Lastly, for B ⊆ A we write A\B for
{a ∈ A|a ∈/ B}.

1. Let R be a commutative ring. By a valuation of R we will mean

a map v : R → G∗ , where G∗ is a totally ordered abelian (possibly
trivial) group which has been extended by ∞, satisfying

Received by the editors on January 26, 1986.

c
Copyright 1990 Rocky Mountain Mathematics Consortium

251
252 K.G. VALENTE

(1) v(xy) = v(x) + v(y),

(2) v(x + y) ≥ min{v(x), v(y)},
(3) v(1) = 0 and v(0) = ∞.
We will, without loss, also assume that G is generated by im(v)\{∞}.
If G is im(v)\{∞}, we then say that v is a Manis valuation, or simply
an M -valuation.
Let v : R → G∗ and w : R → H ∗ be valuations of R. We say that v
and w are equivalent if there exists an order isomorphism θ : G → H
such that w = θ ∗ ◦ v. Let Val(R) (respectively MVal(R)) denote the set
of all (equivalence classes of) valuations (respectively M-valuations) of
R.
For v ∈ Val(R) we know that A(v) := {r ∈ R|v(r) ≥ 0} is a subring
of R and that p(v) := {s ∈ R|v(s) > 0} is a prime ideal of A(v).
We shall refer to these as the valuation subring and valuation prime
associated to v. Lastly we have the infinite ideal associated to v, namely
I(v) := {t ∈ R|v(t) = ∞}.
We note that if v is an M-valuation, then it is completely determined
by the pair (A(v), p(v)). For subrings A and B of R with prime ideals
p and q (respectively) we say (B, q) dominates (A, p) if A ⊆ B and
p = A ∩ q. The following result can be found in [3].

PROPOSITION 1.1 (MANIS). The following are equivalent:

(1) (B, q) = (A(v), p(v)) for some M-valuation v,
(2) r ∈ R\B ⇒ ∃x ∈ q with rx ∈ B\q, and
(3) (B, q) is maximal with respect to domination.

PROPOSITION 1.2. Given v : R → G∗ in Val(R), there is a valuation

v̂ : F (v) → G∗ , where F (v) denotes the field of fractions of R/I(v),
with
v̂(r + I(v)/s + I(v)) = v(r) − v(s)

for all r ∈ R and s ∈ R\I(v).

PROOF. One checks or sees [1].

 A NOTE ON VALUATIONS 253

Let R possess a large Jacobson radical, J(R). That is, for every
r ∈ R, there exists an s ∈ R with r + s ∈ U (R) and rs ∈ J(R). In [2],
J. Grater proved

PROPOSITION 1.3. Let R be as above and A be a valuation subring

of R. Then A is the valuation subring for some M-valuation.

Further, in [6] the author points out that the M-valuation guaranteed
us by this proposition is nothing more than a slight modification of the
original (possibly non-Manis) valuation. Specifically, we have

PROPOSITION 1.4. For R as above and v : R → G∗ a valuation of R,

the mapping v  : R → G∗ defined by
 
 v(r) if r ∈
/ (R : P v)
v (r) =
∞ if r ∈ (R : P v)

is an M-valuation of R.
Clearly the construction described in this proposition gives A(v  ) =
A(v) and p(v  ) = p(v).

2. We now proceed to determine the set of p-primes in a special class

of commutative rings.

LEMMA 2.1. Let F be a field and R a valuation subring of F with

m(R) its maximal ideal. If (B(v), q(v)) is an M-valuation pair of R,
then there exists a valuation subring A of F with B(v) = R ∩ A and
q(v) = R ∩ m(A).

PROOF. Since R is a valuation subring of F , there exists a unique

valuation subring C of F with R ⊆ C and m(C) = I(v). One
now checks that we may naturally identify qf(R/I(v)) with k(C).
With this, there exists a valuation subring A of F with A ⊆ C and
(B(v̂), q(v̂)) = (A, m(A)) (where denotes the image in k(C)). But
the diagram
254 K.G. VALENTE

C k(C) = qf(R/I(v))

φ(I)


R

commutes, where φ(I)(r) = r/1 for all r ∈ R. So

(B(v), q(v)) = (φ−1 (B(v̂)), φ−1 (q(v̂)))

= (R ∩ A, R ∩ m(A)).

For the remainder of this section, R will be a Prüfer domain with

large Jacobson radical and F its quotient field.

LEMMA 2.2. Let R be as above. For (A, m(A)) ∈ Val(F ), the

assignment
(A, m(A)) → (R ∩ A, R ∩ m(A))
provides a well-defined map, ρ, from Val(F ) to MVal(R). Further, this
map is a bijection. We also have


(R 
∩ A, R ∩ m(A)) = (A, m(A)),

where denotes the image under the projection

πI : R  k(RI ) and I = (R : m(A)).

PROOF. For (A, m(A)) ∈ Val(F ), let v (A) be the valuation which is
determined by A. Setting v = v (A)|R , we have v is a valuation of R
and, by Proposition 1.4, B(v) = R ∩ A and q(v) = R ∩ m(A) form an
M-valuation pair of R. Thus our mapping is well-defined.
Continuing with the notation above, we employ a bit of abuse and
let v denote the M-valuation determined by (R ∩ A, R ∩ m(A)). Recall
we know that

v (A)(r), for r ∈
/ (R : m(A)),
v(r) =
∞, for r ∈ (R : m(A)).
 A NOTE ON VALUATIONS 255

Now, letting I = I(v) = (R : m(A)), RI is a valuation subring of F

and k(RI ) is naturally isomorphic to F (v). With πI : R  k(RI )
the natural projection, πI−1 (B(v̂)) is a valuation subring of F with
πI−1 (B(v̂)) ⊆ RI and πI−1 (q(v̂)) its maximal ideal. Also, one can easily
verify that

RI = {r/u|r ∈ R and u ∈ (R ∩ A)\(R ∩ m(A))}.

Claim 1. A = πI−1 (B(v̂)).

Reason. Let r/u ∈ πI−1 (B(v̂)). If r ∈ I, then r ∈ (R : m(A)) ⊆ A.
If this is not the case, then ∞ = v̂(πI (r/u)) = v (A)(r) − v (A)(u) ≥ 0
and again r ∈ A. Thus πI−1 (B(v̂)) ⊆ A.
Now πI−1 (q(v̂)) ⊆ πI−1 (B(v̂)) ∩ m(A). Let s/w ∈ πI−1 (B(v̂)) ∩ m(A)
where s ∈ R and w ∈ U (A) ∩ R. Then s ∈ m(A) and hence
s/w ∈ πI−1 (q(v̂)). Thus (A, m(A)) dominates (πI−1 (B(v̂)), πI−1 (q(v̂)))
and, since the former is a valuation pair, they must in fact be equal.
This establishes the claim. Note that, at this point, we have proved
that

R ∩ A = A and R ∩ m(A) = m(A).
We now wish to show that our map is surjective. To this end let
(B(w), q(w)) ∈ MVal(R). Again we know that I(w) = (R : q(w)) and
(with w denoting the valuation associated to this pair) w determines an
M-valuation ŵ of the valuation subring RI(w) of F . Now, using Lemma
2.1, our assertion quickly follows.
Let A(i) be valuation subrings of F with maximal ideals m(i) for
i = 1, 2. Assume that R ∩ A(1) = R ∩ A(2) and R ∩ m(1) = R ∩ m(2).
Then I(1) = (R : m(1)) = (R : m(2)) = I(2). Further, by our
assumption,

(R 
∩ A(1), R 
∩ m(1)) = (R 
∩ A(2), R ∩ m(2)).

Thus, by the above, (A(1), m(1)) = (A(2), m(2)) ⊆ k(RI(1) ). Hence

(A(1), m(1)) = (A(2), m(2)), and our proof is complete.

THEOREM 2.3. Let R be a Prüfer domain with large Jacobson radical.

Then there is a well-defined map ρ0 : P0 (F ) → P0 (R) with

ρ0 (A, P ) = (R ∩ A, R ∩ P ).
256 K.G. VALENTE

Further, this mapping provides a bijective correspondence between

P0 (F ) and P0 (R).

PROOF. Using Lemma 2.2, we immediately see that (R ∩ A, R ∩ P ) is


a 0-prime of R for any 0-prime (A, P ) of F . Let (R 
∩ A, R ∩ P ) denote
the associated 0-prime of k(RI ) where I = (R : m(A)). From the above
(and continuing with the notation introduced there) we also know that

R ∩ A = A and R  ∩ m(A) = m(A). Moreover we have


R ∩ P = {x/u | x ∈ R ∩ P, u ∈ (R ∩ P )\(R ∩ m(A))}.


Now, with our identifications, R ∩ P = P.
Let (B, T ) be a 0-prime of R with B = (B, q) and let (A, m(A)) be
the unique valuation of F with B = R ∩ A as well as q = R ∩ m(A). By
definition we may think of T as an ordering of A having Supp(T ) :=
T ∩ (−T ) = q. Thus T induces an ordering, T  , of qf(B/q). Now, via
the natural isomorphism from qf(B/q) to k(A), the set

P := {a ∈ A | a + m(A) ∈ T  }

is an ordering of A with Supp(P ) = m(A). Thus (A, P ) is a 0-prime of

F and one checks that R ∩ P = T .
Let (A(i), P (i)) be 0-primes of F for i = 1, 2 with ρ0 (A(1), P (1)) =
ρ0 (A(2), P (2)). By Lemma 2.2 we already know that (A(1), m(1)) =
(A(2), m(2)) = (A, m(A)). Further, by our assumption and the above
remarks, (P (1)) = (P (2)) ⊆ k(A). But this gives P (1) = P (2), and
our map is bijective.

We continue with the notation as above while letting p denote a

rational prime. For (A, P ) ∈ Pp (F ) we shall include it in the subset
Pp∗ (F ) if the following condition is met:

(∗) r∈R and r(R ∩ A) ⊆ P → r ∈ m(A).

For convenience, set B = R ∩ A, q = R ∩ m(A) and Q = R ∩ P . Also,

for specificity in the work that follows, we will employ a new bit of
notation. Namely, if (D, n) is an M-valuation pair of a commutative
 A NOTE ON VALUATIONS 257

ring S, we set F (S, D, n) to be qf(S/(S : n)). As above (D̂, n̂) will

continue to denote the induced valuation pair of F (S, D, n).
From the definition of a p-prime we know that I = I(A, P ) and P
form an M-valuation pair of A with m(A) ⊆ P . Thus, by Proposition
0.4 of [5], I is a valuation subring of F with P its maximal ideal. We
have also seen that F (A, I, P ) may be identified with k(A) and that,
via this identification, Iˆ = πA (I) and P̂ = πA (P ). Since R possesses a
large Jacobson radical, both (B, q) and (I ∩R, Q) are M-valuation pairs
of R and I ∩ R ⊆ B. Hence the latter pair may also be considered as
an M-valuation pair of B. One can check that J := I(B, Q) coincides
with I ∩ R.
By (∗ ), we have (B : Q) = q, and with this F (B, J, Q) ∼ = k(A) via an
isomorphism satisfying (b + q)/(u + q) → bu−1 + m(A) for all b ∈ B
and u ∈ B\q. Hence, up to identification, Jˆ = πA (I), Q̂ = πA (P ) and
ˆ J)
dim(J/p ˆ < ∞ over Z/pZ.
Now, setting L = (R : m(A)), we have RL is a valuation subring of
F and A ⊆ RL . Let denote class modulo the maximal ideal of RL
in k(RL ). Further, we consider three subsets of F (R, B, q):

B̃ := {(b + L)/(u + L) | b ∈ B and u ∈ J\Q},

q̃ := {(x + L)/(u + L) | x ∈ q and u ∈ J\Q}

and
Q̃ := {(y + L)/(u + L) | y ∈ Q and u ∈ J\Q}.
We have B̃ = A and q̃ = m(A) up to the identification of F (R, B, q)
with k(RL ).
258 K.G. VALENTE

We may arrange the information thus far in the following diagram:

R
F

πL
RL

k(RL ) ∼
= F (R, B, q)


(A, m(A)) 
π  π

  L Ā

ρ
(B, q) (A, m(A)) k(A) ∼
= k(A)



(I, P )

π π

L

Ā

ρ
(J, Q) (I, P ) πĀ (πA (I), πA (P ))

LEMMA 2.4. Q̃ = P .

PROOF. Let (y + L)/(u + L) ∈ Q̃. But u ∈ J\Q ⊆ I\P which gives

u−1 ∈ I\P . Hence yu−1 + L ∈ P . With J˜ defined in a manner which
is analogous to the above, we have J˜ = I(B̃, Q̃), (J,
˜ Q̃) and (I, P ) are
both valuations of k(RL ) and J ⊆ I. Clearly Q̃ ⊆ J˜ ∩ P . Also, given
˜
(j + L)/(u + L) ∈ P , we see ju−1 ∈ P but u−1 ∈ / P . By the definition
of a p-prime it follows that j ∈ P . Thus j ∈ Q, Q̃ = J˜ ∩ P , and, by
˜ Q̃) = (I, P ).
maximality, (J,

The valuation (I, P ) ⊆ (A, m(A)) determines a valuation (πĀ (I),

πĀ (P )) of k(A). But we may also identify k(A) with k(A) via
(a+L)+m(A) → a+m(A), and using this we see that (πĀ (I), πĀ (P )) =
(πA (I), πA (P )) as valuations of k(A). Letting G(I) denote the (addi-
tive) value group associated to this valuation, we have

REL(P, n) = G(I)/nG(I) = REL(P , n) = REL(Q̃, n)

 A NOTE ON VALUATIONS 259

for any prime number n (see [5, Proposition 1.5]). Further |REL(Q̃, n)| =
|REL(Q, n)| (see [5, Claim 3 of Theorem 2.3]), and, since (A, P ) is a
p-prime of F , we may conclude that |REL(Q, n)| = n for any prime
number n. Thus we have proved

PROPOSITION 2.5. By defining ρp (A, P ) = (R ∩ A, R ∩ P ), where

R ∩ A = (R ∩ A, R ∩ m(A)), a well-defined mapping, ρp : Pp∗ (F ) →
Pp (R), results.

THEOREM 2.6. The map ρp : Pp∗ (F ) → Pp (R) is a bijection.

PROOF. Fix (B, Q) ∈ Pp (R) with B = (B, q). By Lemma 2.2, there
exists a valuation subring A of F with B = R ∩ A and q = R ∩ m(A).
Also, J = I(B, Q), together with Q, is an M-valuation pair of B, and
(B : Q) = q. As above, F (B, J, Q) ∼
= k(A) so that we may view (J,ˆ Q̂)
as a valuation of k(A). Hence, there exists a unique valuation subring
C of F with C ⊆ A, Jˆ = πA (C) and Q̂ = πA (m(C)).
Now (R ∩ C, R ∩ m(C)) is an M-valuation pair of B. Let j ∈ J. Then
j + m(A) ∈ Jˆ = πA (C). So j ∈ C and J ⊆ R ∩ C. Similarly we have
Q = J ∩ m(C), and, by maximality, (J, Q) = (R ∩ C, R ∩ m(C)).
We consider (A, m(C)). It follows that I(A, m(C)) = C, Ĉ = πA (C)
 = πA (m(C)). So dim(Ĉ/pĈ) = dim(J/p
and m(C) ˆ J)
ˆ < ∞ over Z/pZ.
Also, following the proof above, |REL(Q, n)| = |REL(m(C), n)| = n for
any prime number, n. Thus (A, m(C)) is a p-prime of F which restricts
to (B, Q) in R. By construction we see that (A, m(C)) satisfies (∗ ) and
our map is surjective.
To see that ρp is injective we let (A(i), P (i)) be in Pp∗ (F ), for i = 1, 2,
such that ρp (A(1), P (2)) = ρp (A(2), P (2)). We know immediately
that A(1) = A(2) = A and, letting J = I(R ∩ A, R ∩ P (1)) and
I(i) = I(A, P (i)), for i = 1, 2, R ∩ I(1) = J = R ∩ I(2). Also, via
our past work, πA (I(1)) = Jˆ = πA (I(2)) (as valuation subrings of
k(A)), and, consequently, P (1) = P (2). This completes the proof.

3. We take this opportunity to return to concepts which were intro-

duced in [4]. Specifically we are interested in the holomorphy subring,
260 K.G. VALENTE

H(S), of a semi-real ring S (see §5 of [4 or 6] for appropriate defini-

tions and notation). From the definition, one can immediately verify
that H(S) ⊆ S ∩H(qf(S)) for any domain S. We are particularly inter-
ested in identifying those domains for which the inclusion is equality.
In any such case it easily follows that H(S) would coincide with subring
A(T ) for the preordering T = S ∩ σ(qf(S)) of S.
Assume that R is a Prüfer domain with large Jacobson radical. Here
we have that H(R) = HM (R). Further, by Theorem 2.3, we know that
every 0-prime of R is uniquely determined by a 0-prime of qf(R) via
restriction. Thus it is not surprising that we have (cf. Theorem 5.2
(b))

PROPOSITION 3.1. For R as above and F = qf(R), H(R) = R∩H(F ).

REFERENCES

1. N. Bourbaki, Algébre Commutative, Hermann, Paris, 1964, Chapter 6.

2. J. Grater, Integral closures and valuation rings with zero divisors, Studia Sci.
Math. Hung. 17 (1982), 457 458.
3. M. Manis, Valuations on a commutative ring, Proc. Amer. Math. Soc., 20
(1969), 193 198.
4. K.G. Valente, On orderings, valuations and 0-primes of a commutative ring,
Rocky Mountain J. Math 19 (1989).
5. , The p-primes of a commutative ring, Pacific J. Math 126 (1987),
385 400.
6. , Valuations of a commutative ring: results, counterexamples and
questions, preprint, Center for Pure and Applied Mathematics 376 University of
California, Berkeley, CA. .

DEPARTMENT OF MATHEMATICS, COLGATE UNIVERSITY, HAMILTON, NY 13346

DEPARTMENT OF MATHEMATICS, UNIVERSITY OF CALIFORNIA, BERKELEY, CA
94720