# Integrally closed domain

In commutative algebra, an integrally closed domain A is an integral domain whose integral closure in its field of fractions is A itself. Spelled out, this means that if x is an element of the field of fractions of A which is a root of a monic polynomial with coefficients in A, then x is itself an element of A. Many well-studied domains are integrally closed: fields, the ring of integers Z, unique factorization domains and regular local rings are all integrally closed.

Note that integrally closed domains appear in the following chain of class inclusions:

commutative ringsintegral domainsintegrally closed domainsGCD domainsunique factorization domainsprincipal ideal domainsEuclidean domainsfieldsfinite fields

## Basic properties

Let A be an integrally closed domain with field of fractions K and let L be a field extension of K. Then xL is integral over A if and only if it is algebraic over K and its minimal polynomial over K has coefficients in A. In particular, this means that any element of L integral over A is root of a monic polynomial in A[X] that is irreducible in K[X].

If A is a domain contained in a field K, we can consider the integral closure of A in K (i.e. the set of all elements of K that are integral over A). This integral closure is an integrally closed domain.

Integrally closed domains also play a role in the hypothesis of the Going-down theorem. The theorem states that if AB is an integral extension of domains and A is an integrally closed domain, then the going-down property holds for the extension AB.

## Examples

The following are integrally closed domains.

• A principal ideal domain (in particular: the integers and any field).
• A unique factorization domain (in particular, any polynomial ring over a field, over the integers, or over any unique factorization domain).
• A GCD domain (in particular, any Bézout domain or valuation domain).
• A Dedekind domain.
• A symmetric algebra over a field (since every symmetric algebra is isomorphic to a polynomial ring in several variables over a field).
• Let $\displaystyle{ k }$ be a field of characteristic not 2 and $\displaystyle{ S = k[x_1, \dots, x_n] }$ a polynomial ring over it. If $\displaystyle{ f }$ is a square-free nonconstant polynomial in $\displaystyle{ S }$, then $\displaystyle{ S[y]/(y^2 - f) }$ is an integrally closed domain. In particular, $\displaystyle{ k[x_0, \dots, x_r]/(x_0^2 + \dots + x_r^2) }$ is an integrally closed domain if $\displaystyle{ r \ge 2 }$.

To give a non-example, let k be a field and $\displaystyle{ A = k[t^2, t^3] \subset k[t] }$ (A is the subalgebra generated by t2 and t3.) A is not integrally closed: it has the field of fractions $\displaystyle{ k(t) }$, and the monic polynomial $\displaystyle{ X^2 - t^2 }$ in the variable X has root t which is in the field of fractions but not in A. This is related to the fact that the plane curve $\displaystyle{ Y^2 = X^3 }$ has a singularity at the origin.

Another domain which is not integrally closed is $\displaystyle{ A = \mathbb{Z}[\sqrt{5}] }$; it does not contain the element $\displaystyle{ \frac{\sqrt{5}+1}{2} }$ of its field of fractions, which satisfies the monic polynomial $\displaystyle{ X^2-X-1 = 0 }$.

## Noetherian integrally closed domain

For a noetherian local domain A of dimension one, the following are equivalent.

• A is integrally closed.
• The maximal ideal of A is principal.
• A is a discrete valuation ring (equivalently A is Dedekind.)
• A is a regular local ring.

Let A be a noetherian integral domain. Then A is integrally closed if and only if (i) A is the intersection of all localizations $\displaystyle{ A_\mathfrak{p} }$ over prime ideals $\displaystyle{ \mathfrak{p} }$ of height 1 and (ii) the localization $\displaystyle{ A_\mathfrak{p} }$ at a prime ideal $\displaystyle{ \mathfrak{p} }$ of height 1 is a discrete valuation ring.

A noetherian ring is a Krull domain if and only if it is an integrally closed domain.

In the non-noetherian setting, one has the following: an integral domain is integrally closed if and only if it is the intersection of all valuation rings containing it.

## Normal rings

Authors including Serre, Grothendieck, and Matsumura define a normal ring to be a ring whose localizations at prime ideals are integrally closed domains. Such a ring is necessarily a reduced ring, and this is sometimes included in the definition. In general, if A is a Noetherian ring whose localizations at maximal ideals are all domains, then A is a finite product of domains. In particular if A is a Noetherian, normal ring, then the domains in the product are integrally closed domains. Conversely, any finite product of integrally closed domains is normal. In particular, if $\displaystyle{ \operatorname{Spec}(A) }$ is noetherian, normal and connected, then A is an integrally closed domain. (cf. smooth variety)

Let A be a noetherian ring. Then (Serre's criterion) A is normal if and only if it satisfies the following: for any prime ideal $\displaystyle{ \mathfrak{p} }$,

1. If $\displaystyle{ \mathfrak{p} }$ has height $\displaystyle{ \le 1 }$, then $\displaystyle{ A_\mathfrak{p} }$ is regular (i.e., $\displaystyle{ A_\mathfrak{p} }$ is a discrete valuation ring.)
2. If $\displaystyle{ \mathfrak{p} }$ has height $\displaystyle{ \ge 2 }$, then $\displaystyle{ A_\mathfrak{p} }$ has depth $\displaystyle{ \ge 2 }$.

Item (i) is often phrased as "regular in codimension 1". Note (i) implies that the set of associated primes $\displaystyle{ Ass(A) }$ has no embedded primes, and, when (i) is the case, (ii) means that $\displaystyle{ Ass(A/fA) }$ has no embedded prime for any non-zerodivisor f. In particular, a Cohen-Macaulay ring satisfies (ii). Geometrically, we have the following: if X is a local complete intersection in a nonsingular variety; e.g., X itself is nonsingular, then X is Cohen-Macaulay; i.e., the stalks $\displaystyle{ \mathcal{O}_p }$ of the structure sheaf are Cohen-Macaulay for all prime ideals p. Then we can say: X is normal (i.e., the stalks of its structure sheaf are all normal) if and only if it is regular in codimension 1.

## Completely integrally closed domains

Let A be a domain and K its field of fractions. An element x in K is said to be almost integral over A if the subring A[x] of K generated by A and x is a fractional ideal of A; that is, if there is a $\displaystyle{ d \ne 0 }$ such that $\displaystyle{ d x^n \in A }$ for all $\displaystyle{ n \ge 0 }$. Then A is said to be completely integrally closed if every almost integral element of K is contained in A. A completely integrally closed domain is integrally closed. Conversely, a noetherian integrally closed domain is completely integrally closed.

Assume A is completely integrally closed. Then the formal power series ring $\displaystyle{ AX }$ is completely integrally closed. This is significant since the analog is false for an integrally closed domain: let R be a valuation domain of height at least 2 (which is integrally closed.) Then $\displaystyle{ RX }$ is not integrally closed. Let L be a field extension of K. Then the integral closure of A in L is completely integrally closed.

An integral domain is completely integrally closed if and only if the monoid of divisors of A is a group.

## "Integrally closed" under constructions

The following conditions are equivalent for an integral domain A:

1. A is integrally closed;
2. Ap (the localization of A with respect to p) is integrally closed for every prime ideal p;
3. Am is integrally closed for every maximal ideal m.

1 → 2 results immediately from the preservation of integral closure under localization; 2 → 3 is trivial; 3 → 1 results from the preservation of integral closure under localization, the exactness of localization, and the property that an A-module M is zero if and only if its localization with respect to every maximal ideal is zero.

In contrast, the "integrally closed" does not pass over quotient, for Z[t]/(t2+4) is not integrally closed.

The localization of a completely integrally closed domain need not be completely integrally closed.

A direct limit of integrally closed domains is an integrally closed domain.

## Modules over an integrally closed domain

Let A be a Noetherian integrally closed domain.

An ideal I of A is divisorial if and only if every associated prime of A/I has height one.

Let P denote the set of all prime ideals in A of height one. If T is a finitely generated torsion module, one puts:

$\displaystyle{ \chi(T) = \sum_{p \in P} \operatorname{length}_p(T) p }$,

which makes sense as a formal sum; i.e., a divisor. We write $\displaystyle{ c(d) }$ for the divisor class of d. If $\displaystyle{ F, F' }$ are maximal submodules of M, then $\displaystyle{ c(\chi(M/F)) = c(\chi(M/F')) }$ and $\displaystyle{ c(\chi(M/F)) }$ is denoted (in Bourbaki) by $\displaystyle{ c(M) }$.