NOETHERIAN RINGS

KEITH CONRAD

1. Introduction
In a PID, every ideal has a single generator. In a ring that is not a PID, there may not
be a bound on the number of generators of all the ideals. For example, in the polynomial
ring Q[X, Y ], the ideal (X, Y ) has a generating set of size 2 but not one of size 1 (it is not
principal), and the ideal (X n , X n−1 Y, . . . , XY n−1 , Y n ) in Q[X, Y ] has a generating set of
size n + 1 but not one of size n. Despite these variations, there is an important finiteness
property of the ring Q[X, Y ]: all of its ideals are finitely generated.
Definition 1.1. A commutative ring R is called Noetherian if each ideal in R is finitely
generated.
This name honors Emmy Noether, who in her landmark paper [5] in 1921 proved proper-
ties of such rings by conceptual methods instead of by laborious computations. She referred
to such rings as those satisfying “the finiteness condition” (die Endlichkeitsbedingung). The
label “Noetherian ring” is due to Chevalley [1] in 1943.

2. Examples
A simple (and boring) example of a Noetherian ring is a field. A more general class of
examples are PIDs, since all of their ideals are singly generated. Noetherian rings can be
regarded as a good generalization of PIDs: the property of all ideals being singly generated
is often not preserved under common ring-theoretic constructions (e.g., Z is a PID but
Z[X] is not), but the property of all ideals being finitely generated does remain valid under
many constructions
√ of new rings from old rings. For example, we will see below that every
quadratic ring Z[ d] is Noetherian; many of these rings are not PIDs.
The standard example of a non-Noetherian ring is a polynomial ring K[X1 , X2 , . . . ] in
infinitely many variables over a field K. Non-Noetherian rings need not be “really huge”;
there is a non-Noetherian ring contained in Q[X]: the ring of integral-valued polynomials
Int(Z) = {f ∈ Q[X] : f (Z) ⊂ Z}
X(X−1)
is not Noetherian. This ring is bigger than Z[X], e.g., X2 =


2 is in Int(Z) − Z[X],

X(X−1)···(X−n+1)
as is Xn = n! for all n ≥ 2.

3. Properties of Noetherian rings

Theorem 3.1. The following conditions on a commutative ring R are equivalent:
(1) R is Noetherian: all ideals of R are finitely generated.
(2) each infinite increasing sequence of ideals I1 ⊂ I2 ⊂ I3 ⊂ · · · in R eventually
stabilizes: Ik = Ik+1 for all large k.1
1The notation ⊂ only means containment, not strict containment.
1
2 KEITH CONRAD

(3) Every nonempty collection S of ideals of R contains a maximal element with respect
to inclusion: there’s an ideal in S not strictly contained in another ideal in S.

The first theorem in Noether’s paper [5, p. 30] is that (1) ⇒ (2), and she called this
the “theorem of the finite chain” (Satz von der endlichen Kette). The standard label for
property (2) is the ascending chain condition or ACC. Immediately after proving (1) ⇒ (2),
she observed that (2) ⇒ (1) and therefore that (2) could be used as a definition of her
“finiteness condition” in place of (1).
S
Proof. (1) ⇒ (2): If I1 ⊂ I2 ⊂ · · · is an increasing sequence of ideals, let I = n≥1 In . This
is an ideal since each pair of elements in I lies in a common In , by the increasing condition,
so I is closed under addition and multiplication by elements of R. By (1), I is finitely
generated. Using the increasing condition again, each finite subset of I lies in a common
In , so a finite generating set of I is in some Im . Thus I ⊂ Im , and of course also Im ⊂ I, so
I = Im . Then for all n ≥ m, Im ⊂ In ⊂ I = Im , so In = Im .
(2) ⇒ (1): We prove the contrapositive. Suppose (1) is false, so R has an ideal I that is
not finitely generated. Pick r1 ∈ I. Since I is not finitely generated, I 6= (r1 ), so there is
an r2 ∈ I − (r1 ). Since I 6= (r1 , r2 ), there is an r3 ∈ I − (r1 , r2 ). Proceed in a similar way
to pick rn in I for all n ≥ 1 by making rn ∈ I − (r1 , . . . , rn−1 ) for n ≥ 2. Then we have an
increasing sequence of ideals (r1 ) ⊂ (r1 , r2 ) ⊂ · · · ⊂ (r1 , . . . , rn ) ⊂ · · · in R where each ideal
is strictly contained in the next one, so (2) is false.
(2) ⇒ (3): We will prove the contrapositive. If (3) is false then there is a nonempty
collection S of ideals in R containing no maximal member with respect to inclusion. There-
fore if we start with an ideal I1 in R, we can recursively find ideals I2 , I3 , . . . such that In
strictly contains In−1 for all n ≥ 2. (If there were no ideal in S strictly containing In−1 ,
then In−1 would be a maximal element of S, which doesn’t exist.)
(3) ⇒ (1): Let I be an ideal in R. To prove I is finitely generated, let S be the set of
all finitely generated ideals contained in I. By (3), there is an Ie ∈ S that’s contained in
no other element of S, so Ie is a finitely generated ideal in I and no other finitely generated
ideal of R contains I.e We will show Ie = I by contradiction, which would prove I is finitely
generated. If I 6= I, pick a ∈ I − I.
e e Since Ie is finitely generated, also Ie + Ra is finitely
generated, so I +Ra ∈ S. However, Ie+Ra strictly contains I,
e e which contradicts maximality
of Ie as a member of S. Thus Ie = I. 

The third condition of Theorem 3.1 shows a Noetherian ring R other than the zero ring
has a maximal ideal (let S be the set of proper ideals in R) and every proper ideal I in a
Noetherian ring R is contained in a maximal ideal (let S be the set of proper ideals of R
that contain I). This does not need Zorn’s lemma, which is used to show maximal ideals
exist in arbitrary nonzero commutative rings. Many theorems about general commutative
rings that are proved with Zorn’s lemma can be proved without Zorn’s lemma when the
ring is Noetherian.
The following two theorems put the second condition of Theorem 3.1 to use.

Theorem 3.2. Let R be a Noetherian ring. Each surjective ring homomorphism R → R

is injective, and thus is an isomorphism.

Proof. Let ϕ : R → R be a surjective ring homomorphism. For the nth iterate ϕn (the
n-fold composition of ϕ with itself), let Kn = ker(ϕn ). This is an ideal in R and these
 NOETHERIAN RINGS 3

ideals form an increasing chain:

K1 ⊂ K2 ⊂ K3 ⊂ · · ·
since r ∈ Kn ⇒ ϕn (r) = 0 ⇒ ϕn+1 (r) = ϕ(ϕn (r)) = ϕ(0) = 0, so r ∈ Kn+1 . Since R is
a Noetherian ring, Kn = Kn+1 for some n. Pick r ∈ ker ϕ, so ϕ(r) = 0. The map ϕn is
surjective since ϕ is surjective, so r = ϕn (r0 ) for some r0 ∈ R. Thus 0 = ϕ(r) = ϕ(ϕn (r0 )) =
ϕn+1 (r0 ). Therefore r0 ∈ ker(ϕn+1 ) = ker(ϕn ), so r = ϕn (r0 ) = 0. That shows ker ϕ = {0},
so ϕ is injective. 
The ring R = K[X1 , X2 , . . .] for a field K is not Noetherian and a surjective ring
homomorphism R → R that is not injective is the shift-substitution f (X1 , X2 , . . .) 7→
f (X2 , X3 , . . .) for all f ∈ R.
There is no analogue of Theorem 3.2 for injective ring homomorphisms. For example,
R[X] is a Noetherian ring since it’s a PID and the substitution homomorphism f (X) 7→
f (X 2 ) on R[X] is an injective ring homomorphism that is not surjective.
Theorem 3.3. If R is a Noetherian integral domain that is not a field, then every nonzero
nonunit in R can be factored into irreducibles.
We assume R is not a field because irreducible factorizations don’t have a meaning for
units, so we want R to contain some nonzero elements that aren’t units.
Proof. This will be a proof by contradiction.
Suppose there is an element a in R that is not 0 or a unit and has no irreducible factor-
ization. We will find another nonzero nonunit b ∈ R that does not admit a factorization
into irreducibles and such that there is a strict inclusion of ideals (a) ⊂ (b).
Since a is not irreducible and it is not 0 or a unit, there is a factorization a = bc where b
and c are nonunits (and obviously they are not 0 either). If both b and c have an irreducible
factorization then so does a (just multiply together irreducible factorizations for b and c),
so at least one of b or c has no irreducible factorization. Without loss of generality, say b
has no irreducible factorization. Then since c is not a unit, the inclusion (a) ⊂ (b) is strict.
Rewriting a as a1 and b as a2 , we have a strict containment of ideals
(a1 ) ⊂ (a2 )
where a2 is a nonzero nonunit with no irreducible factorization. Using a2 in the role of a1
in the previous paragraph, there is a strict inclusion of ideals
(a2 ) ⊂ (a3 )
for some nonzero nonunit a3 that has no irreducible factorization. This argument can be
repeated and leads to an infinite increasing chain of (principal) ideals
(3.1) (a1 ) ⊂ (a2 ) ⊂ (a3 ) ⊂ · · ·
where all inclusions are strict. This is impossible in a Noetherian ring, so we have a con-
tradiction. Therefore nonzero nonunits without an irreducible factorization do not exist in
R: all nonzero nonunits in R have an irreducible factorization. 
Theorem 3.3 is not saying a Noetherian integral domain has unique factorization: just
because elements have irreducible factorizations doesn’t mean those are unique (up to the
order of multiplication and multiplication of terms by units). Many Noetherian integral
domains do not have unique factorization.
4 KEITH CONRAD

Remark 3.4. If an integral domain R contains a nonzero nonunit a that has no irreducible
factorization, then Theorem 3.3 tells us R can’t be Noetherian, so R must contain an ideal
that isn’t finitely generated. In fact, the proof of Theorem 3.3 gives us an
S(abstract) example
of an ideal in R that isn’t finitely generated: the union of ideals I := n≥0 (an ) is an ideal
because (an ) ⊂ (an+1 ) for all n, and I isn’t finitely generated because if it were finitely
generated then the containments (an ) ⊂ (an+1 ) could not be strict for all n.
We now show that some basic operations on rings preserve the property of being Noe-
therian.
Theorem 3.5. If R is a Noetherian ring then so is R/I for each ideal I in R.
Proof. Every ideal in R/I has the form J/I for an ideal J of R such that I ⊂ J ⊂ R. Since
R is a Noetherian ring, J is a finitely generated ideal in R, and that finite generating set
for J reduces to a generating set for J/I as an ideal of R/I. 
To create more examples of Noetherian rings we can use the following very important
theorem.
Theorem 3.6 (Hilbert Basis Theorem). If R is a Noetherian ring then so is R[X].
The reason for the name “Basis Theorem” is that a generating set for an ideal may be
called a “basis” even if it’s not linearly independent (cf. the modern term “Gröbner basis”).
The theorem says if each ideal in R has a “finite basis” then this is true of ideals in R[X].
Proof. The theorem is clear if R = 0, so assume R 6= {0}. To prove each ideal I in R[X] is
finitely generated, we assume I is not finitely generated and will get a contradiction.
We have I 6= (0). Define a sequence of polynomials f1 , f2 , . . . in I as follows.
(1) Pick f1 to be an element of I − (0) with minimal degree. (It is not unique.)
(2) Since I 6= (f1 ), as I is not finitely generated, pick f2 in I −(f1 ) with minimal degree.
Note deg f1 ≤ deg f2 by the minimality condition on deg f1 .
(3) For k ≥ 2, if we have defined f1 , . . . , fk in I then I 6= (f1 , . . . , fk ) since I is not
finitely generated, so we may pick fk+1 in I − (f1 , . . . , fk ) with minimal degree.
We have deg fk ≤ deg fk+1 for all k: the case k = 1 was checked before, and for k ≥ 2,
fk and fk+1 are in I − (f1 , . . . , fk−1 ) so deg fk ≤ deg fk+1 by the minimality condition on
deg fk .
For k ≥ 1, let dk = deg fk and ck be the leading coefficient of fk , so dk ≤ dk+1 and
fk (X) = ck X dk + lower-degree terms.
The ideal (c1 , c2 , . . . ) in R (an ideal of leading coefficients) is finitely generated since R
is Noetherian. Each element in this ideal is an R-linear combination of finitely many ck , so
(c1 , c2 , . . . ) = (c1 , . . . , cm ) for some m.
Since cm+1 ∈ (c1 , c2 , . . . , cm ), we have
m
X
(3.2) cm+1 = rk ck
k=1
for some rk ∈ R. From the inequalities dk ≤ dm+1 for k ≤ m, the leading term in fk (X) =
ck X dk + · · · can be moved into degree dm+1 by using fk (X)X dm+1 −dk = ck X dm+1 + · · · , and
this is in I since fk (X) ∈ I and I is an ideal in R[X]. By (3.2), the R-linear combination
Xm
rk fk (X)X dm+1 −dk
k=1
 NOETHERIAN RINGS 5

is in the ideal (f1 , . . . , fm ) and its coefficient of X dm+1 is m

P
k=1 rk ck , which equals the leading
coefficient cm+1 of fm+1 (X) in degree dm+1 . The difference
m
X
(3.3) fm+1 (X) − rk fk (X)X deg fm+1 −deg fk
k=1

is in I, it is not 0 since fm+1 ∈ I − (f1 , . . . , fm ), and it has degree less than dm+1 since
the terms cm+1 X dm+1 cancel out. But fm+1 (X) has minimal degree among polynomials in
I − (f1 , . . . , fm ), and (3.3) is in I − (f1 , . . . , fm ) with lower degree than dm+1 . That’s a
contradiction. Thus I is finitely generated. 

To summarize this proof in a single phrase, “use an ideal of leading coefficients”.

In the proof, the Noetherian property of R is used where we said (c1 , c2 , . . . ) = (c1 , . . . , cm )
for some m. All we need to get the contradiction in the proof is cm+1 ∈ (c1 , . . . , cm ) for
some m. Since (c1 ) ⊂ (c1 , c2 ) ⊂ (c1 , c2 , c3 ) ⊂ · · · , what we need is the following property:
for each infinite increasing sequence of ideals I1 ⊂ I2 ⊂ I3 ⊂ · · · in R, Im = Im+1 for some
m. Of course this is implied by the Noetherian property, but it also implies the Noetherian
property since a non-Noetherian ring has an infinite increasing sequence of ideals with strict
containments at each step: see the proof of (2) ⇒ (1) in Theorem 3.1.

Remark 3.7. Our proof of the Hilbert Basis Theorem, which is due to Sarges [6], is by
contradiction and thus is not constructive. A constructive proof runs as follows. For R 6= 0,
I a nonzero ideal in R[X], and n ≥ 0, let Ln be the set of leading coefficients of polynomials
in I of degree at most n together with 0. This is an ideal in R by the way polynomials
add and get scaled by R. (While Ln might be (0) for small n, Ln 6= (0) for large n since I
contains a nonzero polynomial and multiplying that by powers of X gives us polynomials
in I of all higher degrees.) Since Ln ⊂ Ln+1 , the ideals {Ln } in R stabilize at some point,
say Ln = Lm for n ≥ m. (Thus Lm is generated by the leading coefficients of all nonzero
polynomials in I, so we could have defined Lm that way.) Each Ln has finitely many
generators. When Ln 6= (0), let Pn be a finite set of polynomials of degree at most n in I
whose leading coefficients generate Ln . The union of the finite sets Pn for n ≤ m where
Ln 6= (0) is a generating set for I [4, Sect. 7.10]. This way of proving Hilbert’s basis theorem
is essentially due to Artin, according to van der Waerden [7].

Where in the proof of Theorem 3.6 did we use the assumption that R is Noetherian? It
is how we know the ideals (c1 , . . . , ck ) for k ≥ 1 stabilize for large k, so cm+1 ∈ (c1 , . . . , cm )
for some m. The contradiction we obtain from that really shows cm+1 6∈ (c1 , . . . , cm ) for all
m, so the proof of Theorem 3.6 could be viewed as proving the contrapositive: if R[X] is
not Noetherian then R is not Noetherian.
The converse of Theorem 3.6 is true: if the ring R[X] is Noetherian then so is the ring
R by Theorem 3.5, since R ∼= R[X]/(X)

Corollary 3.8. If R is a Noetherian ring then so is R[X1 , . . . , Xn ].

Proof. We induct on n. The case n = 1 is Theorem 3.6. For n ≥ 2, write R[X1 , . . . , Xn ] as

R[X1 , . . . , Xn−1 ][Xn ], with R[X1 , . . . , Xn−1 ] being Noetherian by the inductive hypothesis,
so we are reduced to the base case. 
6 KEITH CONRAD

Remark 3.9. Corollary 3.8 for R = C was proved by Hilbert in 1890 [3, Theorem 1, p. 474]
as a pure existence theorem in a few pages, not by an algorithm.2 This is what made Hilbert
famous in mathematics. Earlier, Gordan [2] settled the case n = 2 of Corollary 3.8 for R = C
in 1868 by long calculations and spent 20 years unsuccessfully working on n = 3. Hilbert’s
proof for all n was revolutionary, illustrating the power of existence proofs over constructive
methods, which became characteristic of much of modern mathematics. With the rise of fast
computers in the late 20th century, generating sets for polynomial ideals can be computed
routinely with Gröbner bases, which are a multivariable polynomial replacement for the
Euclidean algorithm of polynomials in one variable.
√
Now we can build lots of Noetherian rings. The quadratic ring Z[ d] for a nonsquare
integer d is Noetherian: it’s isomorphic to Z[X]/(X 2 − d), Z[X] is Noetherian √ by Hilbert’s
√
basis√theorem, and Z[X]/(X 2 − d) is Noetherian by Theorem 3.5. Similarly, Z[ 2, 3] and
√
Z[i, 3 2, 7 10] are Noetherian because they are isomorphic to Z[X, Y ]/(X 2 − 2, Y 2 − 3) and
Z[X, Y, Z]/(X 2 +1, Y 3 −2, Z 7 −10). The ring Z[X, 1/X] is Noetherian since it is isomorphic
to Z[X, Y ]/(XY − 1).
For a field K and ideal I in K[X1 , . . . , Xn ], the ring K[X1 , . . . , Xn ]/I is Noetherian since
K is trivially Noetherian. For instance, R[X, Y, Z]/(X 2 + Y 3 − Z 5 , XY Z) is Noetherian.
Remark 3.10. In addition to polynomial rings in finitely many variables, formal power
series rings in finitely many variables are important. For a Noetherian ring R the formal
power series ring R[[X1 , . . . , Xn ]] is Noetherian, and as in the polynomial case writing
R[[X1 , . . . , Xn ]] as R[[X1 , . . . , Xn−1 ]][[Xn ]] reduces the proof to the case n = 1. A formal
power series usually doesn’t have a leading coefficient, so the proof in the polynomial case
doesn’t work directly for power series. What can be used with formal power series instead of
a leading term is a lowest degree term, so the proof of Theorem 3.5 can be adapted to formal
power series by changing highest-degree coefficients into lowest-degree coefficients, although
an infinite “limiting process” occurs in the proof since the multipliers on a generating set
for the ideal will be power series. See [4, Theorem 7.11].

References
[1] C. Chevalley, On the Theory of Local Rings, Ann. of Math. 44 (1943), 690–708. Online at https://
www.jstor.org/stable/1969105.
[2] P. Gordan, Beweis, dass jede Covariante und Invariante einer binären Form eine ganze Function mit
numerischen Coefficienten einer endlichen Anzahl solcher Formen ist, J. Reine Angew. Mathematik 69
(1868), 323–354. Online at https://eudml.org/doc/148066.
[3] D. Hilbert, Ueber die Theorie der algebraischen Formen, Math. Annalen 36 (1890), 473–534. Online at
https://eudml.org/doc/157506.
[4] N. Jacobson, “Basic Algebra II”, 2nd ed., W. H. Freeman & Co., New York, 1989.
[5] E. Noether, Idealtheorie in Ringbereichen, Math. Annalen 83 (1921), 24–66. Online at https://
eudml.org/doc/158855. English translation by D. Berlyne https://arxiv.org/abs/1401.2577.
[6] H. Sarges, Ein Beweis des Hilbertschen Basissatzes, J. Reine Angew. Math. 283 (1976), 436–437. Online
at https://eudml.org/doc/151744.
[7] B. L. van der Waerden, On the sources of my book Moderne Algebra, Historia. Math. 2 (1975), 31–40.
Online at https://core.ac.uk/download/pdf/82253306.pdf.

2Hilbert could not use the proof that we gave for his basis theorem, since he didn’t have the concept of
a Noetherian ring in full generality available to him.