Noetherianity of polynomial rings up to group actions

The conclusion follows from [MM, Theorem III.9.1] and the above argument.

Suppose that \mathcal{A} is finitely generated. By Proposition 2.1, one can assume that

whose elements are equivalence classes. Given an equivalence class [a] \in A, one can find a representative a \in \mathcal{A}(G/H) for a certain H \in \mathcal{H}. Since \mathcal{A} is finitely generated, we can find a finite set \{a_1, \, \ldots, \, a_n\} with a_i \in \mathcal{A}(G/H_i) and H_i \in \mathcal{H} which generates \mathcal{A}. In particular, a can be written as a polynomial of elements \mathcal{A} (\sigma_{i,j}) (a_i) \in \mathcal{A}(G/H) for finitely many \sigma_{i, j} \in \mathcal{O}_G(G/H, \, G/H_i). But for each \sigma_{i, j} one can find an element g_{i, j} \in G such that g_{i, j} \cdot [a_i] = [\mathcal{A} (\sigma_{i, j}) (a_i)]. Consequently, [a] can be written as a polynomial of elements g_{i, j} \cdot [a_i] \in A. Thus A is generated by elements lying in the orbits containing [a_i], 1 \leqslant i \leqslant n.

Conversely, suppose that A is finitely generated up to G-action. Then one can find a finite set \{a_1, \, \ldots, \, a_n\} such that A is generated by elements lying in the G-orbits G\cdot a_i, 1 \leqslant i \leqslant n. Again by Proposition 2.1, one can assume that \mathcal{A}(G/H) = A^H for H \in \mathcal{H}. For each a_i, since \operatorname{Stab}_G(a_i) is an open subgroup of G, one can find a certain H_i \in \mathcal{H} such that H_i \leqslant \operatorname{Stab}_G(a_i). Then a_i \in \mathcal{A}(G/H_i). Let \mathcal{B} be a subsheaf (of commutative rings) of \mathcal{A} such that a_i \in \mathcal{B}(G/H_i) for 1 \leqslant i \leqslant n, and let B = \psi(\mathcal{B}) which can be regarded as a subring of A. Then a_i \in \mathcal{B} (G/H) = B^H \subseteq B. Consequently, \{a_1, \ldots, a_n \} \subseteq B, which forces A = B. Consequently, \mathcal{B} = \mathcal{A}, so \mathcal{A} is finitely generated.

Each k[G/H_i] is generated by one element up to G-action, namely the indeterminate corresponded to H \in G/H, so the left side is finitely generated up to G-action. It is easy to check that its quotient algebras are also finitely generated up to G-action.

Conversely, if A is finitely generated up to G-action, then one can find elements a_i \in A, 1 \leqslant i \leqslant n, such that A is generated by elements in the G-orbits G \cdot a_i. For each a_i, let H_i be a member in \mathcal{H} which is also a subgroup of \operatorname{Stab}_G(a_i). Then one can construct a k-algebra homomorphism k[G/H_i] \to A by sending the indeterminate corresponded to H_i \in G/H_i to a_i. This map is well defined by the definition of H_i. Since the image of the map specified in the proposition contains each a_i, it coincides with A. Thus the map is surjective.

The equivalence between (1) and (3) can be established as Lemma 3.1, based on Theorem 2.2, so we only need to show the equivalence between (1) and (2). Note that V is a finitely generated AG-module if and only if there exists a finite set \{ v_1, \ldots, v_n \}\subseteq V such that each v \in V can be written as a finite linear combination \lambda_1 v_1 + \ldots + \lambda_n v_n with \lambda_i \in AG. But each \lambda_i can be expressed as a finite linear combination a_{i1}g_1 + \ldots + a_{is_i} g_{s_i}. Consequently, statement (1) holds if and only if each v \in V can be written as a finite linear combination of elements in G \cdot v_1 \cup \ldots \cup G \cdot v_n with coefficients in A, which is precisely statement (3).

Let V be a discrete AG-module. For v \in V, take H_v in \mathcal{H} such that H_v \leqslant \operatorname{Stab}_G(v). Since there is a homomorphism A(G/H_v) \to V sending H_v \in G/H_v to v, we can define a surjection

establishing the first statement. The second statement follows immediately.

The first statement follows from Theorem 2.2 and Lemma 3.5. The second statement is also clear since G-invariant ideals of A are precisely discrete AG-submodules of A. The third statement follows from Proposition 3.7.

The if direction is clear since given two entrywise distinct sequences (s_1, \, s_2, \, \ldots, \, s_n) and (t_1, \, t_2, \, \ldots, \, t_n), the injection sending s_i to t_i can be extended to a certain \sigma \in G. For the other direction, note that an injection \sigma: L \to T can be identified with an entrywise distinct sequence

whose entries exhaust all elements in T. Then any element \tilde{\sigma} \in G sending (l_1, \, \ldots, \, l_m, \, t_{m+1}, \, \ldots, \, t_n) to the above sequence extends \sigma.

Now we prove the second statement. If the map is not injective, then there are finite subsets K \neq L such that H_K = H_L, so one has K \neq K \cup L but H_K = H_{K \cup L} or L \neq K \cup L but H_L = H_{K \cup L}. Consequently, one gets a finite set T and a proper subset T' \subsetneqq T such that H_T = H_{T'}. Write T as a sequence

such that the first m elements form T', and define another sequence \mathbf{t}' via replacing t_{m+n} by another element not in T. Then there is no \sigma \in G sending the sequence \mathbf{t} to \mathbf{t}' since if \sigma fixes the first m terms, it must fix the whole sequence. Therefore, the action of G on S is not highly transitive.

(1) \Rightarrow (2). Given two entrywise distinct sequences \mathbf{t} = (t_1, \, \ldots, \, t_n) and \mathbf{t}' = (t_1', \, \ldots, \, t_n'), and let T and T' be the corresponded finite subsets of S. Since T and T' are isomorphic in {\mathrm{FI}}^{{\mathrm{op}}}_S, so must be G/H_T and G/H_{T'} in \mathcal{O}_G. In particular, there is g \in G such that gH_Tg^{-1} = H_{T'}, so g(t) = \sigma g(t) for all t \in T and \sigma \in H_{T'}. It follows that g(t) \in T' since otherwise every element in H_{T'} fixes T'' = T' \sqcup \{g(t) \} and hence H_{T'} = H_{T''}, contradicting the assumption that the map T \mapsto G/H_T is an bijection.

We have deduced that every g \in G representing an isomorphism from G/H_T to G/H_{T'} gives a bijection from T to T' via restricting g to T'. Furthermore, by the definition of orbit categories we know that g and g' in G representing the same isomorphism from G/H_T to G/H_{T'} if and only if H_{T'} g = H_{T'} g', and if and only if their restrictions coincide as maps from T to T'. Since \varrho is an isomorphism, the number of morphisms from G/H_T to G/H_{T'} equals the number of bijections from T' to T. It follows from these observations that there is a certain g \in G representing a morphism from G/H_T to G/H_{T'} whose restriction to T' sends \mathbf{t}' to \mathbf{t}; that is, the action of G on S is highly transitive.

(2) \Rightarrow (1). The proof is similar to that of [MM, Theorem III.9.2]. For the convenience of the reader, we describe an explicit construction of \varrho: {\mathrm{FI}}_S^{{\mathrm{op}}} \to \mathcal{O}_G which is fully faithful and is a bijection between the object sets. Since the map sending a finite subset T \subseteq S to G/H_T is the desired bijection by Lemma 4.3, we only need to define \varrho for morphisms.

Given a morphism \varsigma: T \to L in {\mathrm{FI}}_S^{{\mathrm{op}}}, we obtain a unique injection \sigma: L \to T. By Lemma 4.3, \sigma extends to an element \tilde{\sigma} \in G. Since \tilde{\sigma}^{-1}H_T \tilde{\sigma} is contained in H_L, \tilde{\sigma}^{-1} represents a morphism from G/H_T to G/H_L in \mathcal{O}_G sending gH_T to g \tilde{\sigma} H_L, so we can define \varrho(\varsigma) to be this morphism.

We check that the above construction is well defined. Indeed, if \tilde{\sigma} and \tilde{\tau} are extensions of \sigma: L \to T and \tau: L \to T respectively, then for any g \in G,

Thus this construction is independent of the choice of extensions, and hence is well defined. It also shows that the construction is faithful. It is routine to check that \varrho is indeed a functor, and it is full since every morphism in {\mathrm{FI}}_S extends to an element in G and every morphism in \mathcal{O}_G is induced by an element in G.

(2) \Leftrightarrow (3). This is well known to experts; see for instance [Ca, Section 4.2]. For the convenience of the reader we give a detailed proof. Without loss of generality we can assume that \rho is an injective homomorphism, and in this case the image of \rho is precisely G. Note that the collection \{ \sigma U_T \}, where \sigma \in \operatorname{Sym}(S) and U_T is the stabilizer subgroup of \operatorname{Sym}(S) fixing every element in a finite subset T \subseteq S, is a base of the topology on \operatorname{Sym}(S). Therefore, G is dense in \operatorname{Sym}(S) if and only if G \cap \sigma U_T is nonempty for every finite subset T \subseteq S and every \sigma \in \operatorname{Sym}(S); that is, there is a certain g \in G such that g \in \sigma U_T. But this is equivalent to saying that for every \sigma \in \operatorname{Sym}(S) and every finite subset T \subseteq S, there is a certain g \in G such that g and \sigma induce the same map from T to \sigma(T). But the action of \operatorname{Sym}(S) on S is highly transitive, so this is true if and only if the action of G on S is highly transitive.

The conclusion follows from Theorem 2.2 and Proposition 4.4.

By Proposition 4.4, the orbit category \mathcal{O}_G is isomorphic to the opposite category of {\mathrm{FI}}_S. Let \mathcal{A} be the structure sheaf over the site (\mathcal{O}_G, J_{at}) corresponded to A = k[S]. By Example 3.4, for finite subsets \mathcal{T} of S,

and for an injection \sigma: L \to T, \mathcal{A}(\sigma) is the inclusion

In other words, \mathcal{A} is the structure sheaf \mathbf{X}^{{\mathrm{FI}}, 1} in [NR1, Definition 2.17].

Now let V be a finitely generated discrete AG-module. By Proposition 3.9, it suffices to show the Noetherianity of the corresponded \mathcal{A}-module \mathcal{V}. But by [NR1, Theorem 6.15], \mathcal{V} as a presheaf of modules is Noetherian, which automatically implies the Noetherianity of \mathcal{V} as an \mathcal{A}-module. Indeed, given a \mathcal{A}-submodule \mathcal{W} of \mathcal{V}, we can viewed \mathcal{W} as a subpresheaf of \mathcal{V}. Since \mathcal{V} is Noetherian as a presheaf of modules, \mathcal{W} as a presheaf is finitely generated. Therefore, by [NR1, Proposition 3.18], we obtain a surjection

of presheaves of modules. But by Example 3.8 \mathbf{F}^{{\mathrm{FI}}, d_i} is the \mathcal{A}-module corresponded to the discrete AG-module A(G/H_{T_i}) with |T_i| = d_i, so this map is also a surjection of \mathcal{A}-modules. It follows from Lemma 3.5 and a sheaf theoretic version of Proposition 3.7 that \mathcal{W} is a finitely generated \mathcal{A}-module, so \mathcal{V} is Noetherian.

(1). If (S, \leqslant) has an endpoint, say, a minimal element a. Then for any b \neq a in S, the open interval (-, a) is empty, while (-, b) is nonempty, so they cannot be isomorphic. Thus (S, \leqslant) is unbounded. Similarly, if S is not dense, then we can find an increasing pair a < b in S with (a, b) = \emptyset. Since S is unbounded, we can find another c \in S with c > b. Then (a, b) is not isomorphic to (a, c). This contradiction tells us that (S, \leqslant) is dense.

(2). The only if direction is clear. For the other direction, since \leqslant is unbounded, we can find a certain c \in S such that c > a and c > a'. Then one has (a, c) \cong (a', c) and hence

Similarly, one can construct a desired isomorphism (-, a) \cong (-, a').

(3). The only if direction is clear. For the if direction, one needs to show (-, a) \cong (S, \leqslant) \cong (a, -) for a \in S. We prove the first isomorphism since the second one can be verified similarly.

We claim that \leqslant is unbounded. Otherwise, suppose that s_0 is the minimal element in S. For any s < t in S, let \sigma: (s, t) \to (S, \leqslant) be an isomorphism. Then one can find a certain s' \in (s, t) such that \sigma(s') = s_0. Consequently, one has s' > s and s' is the minimal element in (s, t). But then (s, s') is the empty set, which cannot be isomorphic to (S, \leqslant). This contradiction tells us that S has no minimal element. Similarly, one can show that it has no maximal element.

Choose an element x \in (s, t). Then the isomorphism \sigma: (s, t) \to (S, \leqslant) induces an isomorphism (s, x) \cong (-, \sigma(x)). But (s, x) \cong (S, \leqslant) as well, so it remains to show that (-, a) \cong (-, \sigma(x)). But this is clear since by statement (2), \leqslant is homogenous.

(1) \Rightarrow (4). Let \sigma: T \to L be an order-preserving injection. Without loss of generality we can assume that T \neq \emptyset. Order elements in T as a_1 < \ldots < a_m and elements in L as b_1 < \ldots < b_n with m \leqslant n. By Lemma 5.2, one can choose a certain s, t \in S such that s < a_1 < a_m < t and s < b_1 < b_n < t. A desired order-preserving permutation \tilde{\sigma} can be constructed via glueing \sigma, the following isomorphisms

and identities on (-, s) and (t, -).

(4) \Rightarrow (3). For n \in \mathbb{N}, a_1 < \ldots < a_n, and b_1 < \ldots < b_n, let L and T be the subsets of S consisting of a_i’s and b_i’s respectively. Then there is a unique order-preserving bijection \sigma: L \to T extending to an order-preserving bijection \tilde{\sigma} \in \operatorname{Aut}(S, \leqslant). Consequently, the action of \operatorname{Aut}(S, \leqslant) on (S, \leqslant) is highly homogenous.

(3) \Rightarrow (2) and (4) \Rightarrow (5) are clear.

(2) \Rightarrow (1). Take a < b and a' < b' in S and choose an element \sigma in the group such that \sigma(a) = a' and \sigma(b) = b'. Clearly, \sigma induces the desired isomorphisms (-, a) \cong (-, a'), (a, b) \cong (a', b'), and (b, -) \cong (b', -).

(5) \Rightarrow (4). Given an order-preserving injection \sigma from a finite subset T of S to another finite subset L of S, we obtain an order-preserving bijection \sigma' from T to \sigma(T). By (5), \sigma' extends to an order-preserving bijection \tilde{\sigma}: S \to S, which clearly extends \sigma.

The only if direction. Let t_1 < \ldots < t_n and t_1' < \ldots < t'_n be two increasing sequences in S, and let T (resp., T') be the set formed by elements in the first (resp., second) sequences. Since T is isomorphic to T' in {\mathrm{OI}}_S^{{\mathrm{op}}}, G/H_T and G/H_{T'} are isomorphic in \mathcal{O}_G. Thus we can find a certain \sigma \in G such that \sigma H_T \sigma^{-1} = H_{T'}. As in the proof of Proposition 4.4, one can show that \sigma is an extension of the unique increasing map from T' to T. Thus the action of G on (S, \leqslant) is highly homogenous.

The if direction. Firstly we show that the map T \mapsto H_T is a bijection. Since it is clearly surjective, it suffices to show the injectivity. Suppose that T \neq L are two finite subsets of S and without loss of generality assume that there is an element x contained in T but not contained in L. Order elements in T \cup L as a sequence x_1 < x_2 < \ldots < x_n and suppose that x = x_i. Note that the linear order \leqslant is homogeneous by Lemma 5.5, and hence dense by Lemma 5.2, so we can replace x_i by another element y not contained in this sequence to obtain a new increasing sequence. Again by Lemma 5.5 we can find a certain \sigma \in G sending the first sequence to the second sequence. Clearly, \sigma is contained in H_L but not in H_T, establishing the injectivity of the map T \mapsto H_T. The desired functor \varrho can be constructed by slightly modifying the proof of Proposition 4.4.

The conclusion follows from Theorem 2.2 and Proposition 5.7.

Applying the same argument as that in the proof of Theorem 4.6, one gets the following description of the structure sheaf \mathcal{A} over the site ({\mathrm{OI}}_S, J_{at}) corresponded to the algebra A = k[S]: for a finite subset T of S,

for an order-preserving injection \alpha: L \to T, \mathcal{A}(\alpha) is the inclusion

In other words, \mathcal{A} coincides with the structure sheaf \mathbf{X}^{{\mathrm{OI}}, 1} in [NR1, Definition 2.17].

Let V be a finitely generated discrete AG-module. By Proposition 3.9, it suffices to show that the corresponded \mathcal{A}-module \mathcal{V} is Noetherian. But this has been established in [NR1, Theorem 6.15]. Indeed, the conclusion proved [NR1] is that \mathcal{V} as a presheaf of modules is Noetherian, automatically implying the Noetherianity of \mathcal{V} as an \mathcal{A}-module.

Let V be a k[S]\mathcal{I}-submodule of k[S] and G = \operatorname{Aut}(S, \leqslant). Then V viewed as a k[S]G-submodule of k[S] shall be finitely generated by Theorem 5.10. Since G is a submonoid of \mathcal{I}, it follows that V as a k[S] \mathcal{I}-module is finitely generated as well. Therefore, k[S] is a Noetherian k[S]\mathcal{I}-module.

The proof is based on a very transparent observation: if H \leqslant K, then K-invariant ideals of k[S] form a subposet of the poset of H-invariant ideals of k[S]. Consequently, if k[S] is Noetherian up to the action of H, it is also Noetherian up to the action of K. Therefore, it is enough to show that \operatorname{Aut}(S, \leqslant) is a subgroup of G, namely every order-preserving bijection is a self-homeomorphism on (S, \mathcal{T}_{\mathrm{ord}}). But this is trivially true.

Without loss of generality we can assume that G is a subgroup of \operatorname{Sym}(S). We have established the conclusion for dense subgroups of \operatorname{Sym}(S) and \operatorname{Aut}(S, \leqslant) in Theorem 4.5 and Theorem 5.10. Thus it remains to check the conclusion for dense subgroups of \operatorname{Aut}(S, \mathcal{B}), \operatorname{Aut}(S, \mathcal{K}), or \operatorname{Aut}(S, \mathcal{S}). But as we did before, we can show the following fact: If G is a dense subgroup of \operatorname{Aut}(S, \mathcal{B}), then its orbit category is equivalent to the category of finite sets of S equipped with linear betweenness relations and structure-preserving injections. Therefore, by the downward Löwenheim–Skolem theorem (see [Marker, Section 2.3]), Theorem 2.2, and the well known fact that this structure is \aleph_0-categorical (see [Ca, Section 2.5, (2.10)]), the conclusion is independent of particular choices of sets S, linear betweenness relations on S, and dense subgroups G \leqslant \operatorname{Aut}(S, \mathcal{B}). Similarly, one can prove the independence result for \mathcal{K} or \mathcal{S}. Consequently, we only need to check the conclusion for the following three groups:

• G_1 = \operatorname{Aut}(\mathbb{Q}, \mathcal{B}),

• G_2 = \operatorname{Aut}(\mathbb{Q}, \mathcal{K}),

• G_3 = \operatorname{Aut}(\mathbb{Q}, \mathcal{S}),

where \mathcal{B}, \mathcal{K} and \mathcal{S} are induced from the natural linear order on \mathbb{Q}. For details, see [BMMN, Section 11.3].

By [BMMN, Theorems 11.7], G_1 contains \operatorname{Aut}(\mathbb{Q}, \leqslant) as a normal subgroup of index 2, so the conclusion holds for G_1. By [BMMN, Theorem 11.11], G_3 contains G_2 as a normal subgroup of index 2, so it is enough to prove the conclusion for G_2. But by [BMMN, Theorem 11.9], G_2 contains H = \operatorname{Aut}(\mathbb{Q}_0, \leqslant) as a subgroup with \mathbb{Q}_0 = \mathbb{Q} \setminus \{0 \}. Note that (\mathbb{Q}_0, \leqslant) is isomorphic to (\mathbb{Q}, \leqslant). Thus by Theorem 5.10, k[\mathbb{Q}] = k[x_0][\mathbb{Q}_0] is Noetherian under the action of H, so it is also Noetherian under the highly homogenous action of G_2 on \mathbb{Q}_0. This finishes the proof.

Since the property that every nontrivial open interval is isomorphic to the whole set is not a first order sentence, instead we prove a stronger conclusion: every infinite set S can be equipped with a linearly ordered field structure (S, +, \cdot, 0, 1, <, h) satisfying the additional global homogeneity: for every a < b and c < d in S, there exists p, q \in S such that x \mapsto px+q is an order-preserving bijection from (a, b) to (c, d); h is an order-preserving bijection from (0, 1) to S. Clearly, this strengthened global homogeneity is first order definable. Hence, every structure of above theory is also a globally homogenous linear order. Since, for some suitable h, (\mathbb{Q}, +, \cdot, <, h) is a countable model of this theory by Cantor’s theorem, the desired conclusion follows from Löwenheim–Skolem theorem (see [Marker, Section 2.3]).

Suppose that the axiom of choice holds, and let S be an infinite set with cardinality \kappa. By Lemma 5.18, one can construct a globally homogenous linear order \leqslant on another set S' with the same cardinality. Then one can define a linear order on S via an arbitrary bijection from S to S', which is clearly globally homogenous.

Now suppose that every infinite set can be equipped with a globally homogeneous linear order. We want to to show that every infinite set admits a well order, which is equivalent to the axiom of choice. Fix an infinite set X, and set \kappa = \aleph(P(X)) where P(X) is the power set of X. Without loss of generality, we may assume that X \cap \kappa = \emptyset.

Fix a globally homogeneous linear order \leqslant on X \cup \kappa, and define a map f: \kappa \rightarrow P(X) by

By the definition of \kappa, f cannot be an injection, so there are \alpha < \beta such that f(\alpha) = f(\beta). This happens if and only if the interval (\alpha, \beta) is disjoint from X, so (\alpha, \beta) \subseteq \kappa. Since \kappa is an ordinal, it is well orderable, so is (\alpha, \beta). But since \leqslant is globally homogeneous, one can find an isomorphism \rho: (\alpha, \beta) \to (X \cup \kappa, \leqslant). Using this bijection and a well order \preccurlyeq on (\alpha, \beta), one can define a well order on X \cup \kappa, which induces a well order on X.