Why a chain condition matters
A submodule can be described by generators even when no basis is available. The difficult question is whether an arbitrary collection of relations can be reduced to finitely many. A Noetherian condition gives a precise answer: every submodule has a finite generating set. It builds on the preceding discussion of modules and exact sequences.
Throughout, is a commutative ring with identity and modules are unital. We allow the zero ring and the zero module. A finitely generated module need not itself be Noetherian unless its ring has an additional property; the distinction will be proved rather than assumed.
Two ways to say Noetherian
Definition 8.1.1
(Noetherian modules and the ascending chain condition)
An -module is Noetherian if every submodule of is finitely generated. An ascending chain of submodules is a sequence . It stabilizes if there is some with for every . The ascending chain condition (ACC) says every such chain stabilizes.
Theorem 8.1.2
(Finite generation and ascending chains)
An -module is Noetherian if and only if its submodules satisfy ACC.
Proof
(Unions of chains and successive choices)
Suppose every submodule is finitely generated and take an ascending chain. Its union is a submodule: any two members lie together in a sufficiently late term. Choose finitely many generators of . All lie in one term , so and the chain stabilizes.
Conversely, suppose ACC holds and a submodule is not finitely generated. Choose ; after choosing , their span is a proper submodule of , so choose outside it. The spans form a strictly ascending chain, contrary to ACC. Therefore every submodule is finitely generated. The inductive choices use dependent choice, available in standard ZFC foundations.
Exact sequences preserve the condition
Proposition 8.1.3
(Submodule and quotient test)
For a submodule , the module is Noetherian if and only if both and are Noetherian.
Proof
(Generators of the intersection and of the image)
If is Noetherian, every submodule of is a submodule of . A submodule of is the image of its preimage under the quotient map; that preimage is finitely generated, so its image is too.
For the converse take any . The submodule is generated by finitely many , and the image is generated by finitely many cosets with . For any , subtract a linear combination of the representing its coset. The remainder lies in and is a combination of the . Thus the combined finite list generates . This works without a splitting or a choice of basis.
A ring is Noetherian when it is Noetherian as a module over itself, equivalently every ideal is finitely generated. If is Noetherian, is Noetherian by induction using the exact sequence . Every finitely generated -module is a quotient of some , hence Noetherian. The converse is not true for an arbitrary finitely generated module over an arbitrary ring: itself is cyclic even when is not Noetherian.
Adding a polynomial variable
Theorem 8.1.4
(Hilbert basis theorem)
If is Noetherian, then is Noetherian. Consequently is Noetherian for every finite .
Proof
(Leading coefficients and degree reduction)
Let be an ideal of . The leading coefficients of its nonzero polynomials, together with zero, form an ideal of : multiplying by a scalar and adding two leading coefficients can be implemented at a common degree by multiplying the corresponding polynomials by powers of ; cancellation of the common leading terms merely gives the zero sum. For scalar multiplication, is either zero or the leading coefficient of , even when has zero divisors. If then and there is nothing further to prove. Otherwise, since is Noetherian, choose nonzero generators of and polynomials with leading coefficient .
Choose at least every . Let be the -module of polynomials of degree less than , with . It is a finite free -module, so is finitely generated as an -module; choose generators .
For of degree , write its leading coefficient as . Subtract . The result stays in and has strictly smaller degree, or is zero. Repeating finitely many times leaves an element of . Therefore the finitely many and generate as an ideal. No division by a leading coefficient is used. Repeating the one-variable result proves the finite-variable statement.
Proof X-ray
Two uses of the Noetherian hypothesis
The hypothesis that is Noetherian is used twice: once to choose finitely many generators of the leading-coefficient ideal , and once to make finitely generated as a submodule of the finite free module . The key move is the subtraction of : it removes the leading term by writing the leading coefficient as a combination of the , so no leading coefficient is ever inverted and zero divisors cause no trouble. The process terminates because each step lowers the degree, and it closes because every polynomial subtracted lies in , so the final remainder lies in . Dropping the would lose exactly this remainder.
Examples and limits
Worked example 8.1.5
(A polynomial ideal with a concrete generating set)
Over a field , let . Since , we have . The displayed reduction works in every characteristic; the conclusion comes from explicit combinations, not from assuming divides every generator.
Worked example 8.1.6
(A cyclic module over a non-Noetherian ring)
Let , polynomials in countably many variables with each polynomial using only finitely many variables. The ideals form a strict chain: setting to zero and to one shows is not in the th ideal. Thus is not Noetherian, although the -module has the single generator . The finite-variable conclusion of the Hilbert basis theorem cannot be extended by simply writing an infinite list of variables.
Worked example 8.1.7
(Recovering a low-degree generator in )
Let . Its leading-coefficient ideal is , because has leading coefficient ; so we may take , and . The low-degree part consists of the constants in . Writing a constant as and setting gives , so , with generator . For , subtract to get , then subtract to get . Hence
Here alone generates only , which does not contain : the constant generator comes from the bounded-degree part, not from the leading coefficients.
Common mistake
(Keeping only the leading-coefficient lifts)
It is tempting to stop once the leading coefficients of the generate . The previous example shows why this fails: is generated by the leading coefficient of , yet . Two ideals with the same leading-coefficient ideal can differ in low degrees, so the generators of are an essential part of the proof.
Counterexample mode
Ascending and descending are different
The -module is Noetherian because every ideal is generated by one nonnegative integer. Yet never stabilizes. Noetherian does not mean that descending chains stabilize; that is a separate Artinian condition.
Exercises
Quick check
(Prove that every quotient of a Noetherian module is Noetherian.)
Solution · Answer
If and , then is a finitely generated submodule of . The images of its generators generate .
Quick check
(Show that is Noetherian as a -module.)
Solution
The ring is Noetherian because its ideals are principal. The quotient theorem applies to . Alternatively, the module is finite, so it has only finitely many submodules.
Quick check
(Let be a field. Prove is Noetherian, and explain why this does not say every ideal is generated by one polynomial.)
Solution
A field has only the ideals zero and the whole field, so it is Noetherian. Apply the Hilbert basis theorem twice. For a counterexample to principality, is not principal: a common divisor of and in must be a unit, but is proper because evaluation at annihilates it.
Quick check
(In , prove the ideal generated by all variables is not finitely generated.)
Solution
If finitely many polynomials generated that ideal, their variables would lie in some finite set . Each generator has zero constant term. Setting those variables to zero but makes every generator vanish and nonzero, a contradiction.
Further reading
The Stacks Project, Noetherian rings gives a finite-type form of the Hilbert basis theorem, and More Noetherian rings states the corresponding finite-module and ascending-chain facts.