Proposition

Let kk be a field, with fk[x]f \in k[x] a non-constant polynomial. Let RR be a subring of k[x]k[x] such that fRf \in R and kR.k \subseteq R. Define the map φ:k[t]R\varphi : k[t] \to R to be constant on elements of kk and send tt to f.f. This maps gives every such RR a k[t]k[t]-algebra structure. Then

  1. k[x]k[x] is finitely generated as a k[t]k[t]-module
  2. every subring RR (as above) is finitely generated as a k[t]k[t]-module
  3. every subring of k[x]k[x] containing kk is a Noetherian ring.

Proof

1.

Take the set S={xnnN,n<deg(f)}.S = \{ x^n \mid n \in \mathbb{N}, n < \deg(f) \}. Any other power of xx can be written by taking a sum of elements of SS scaled by an appropriate elements of k[t]k[t] using the following procedure.

For each nNn \in \mathbb{N} we can write it down as m1deg(f)+m2m_1\deg(f) + m_2 with m1,m2Nm_1,m_2 \in \mathbb{N} and m2<deg(f).m_2 < \operatorname{deg}(f). Define gng_n as φ(tm1/cm1)xm2\varphi(t^{m_1}/c^{m_1})x^{m_2} with cc being the leading coefficient of f.f. It is clear that gng_n is a monic polynomial of degree n.n. Moreover, since xm2S,x^{m_2} \in S, this doesn’t go beyond our generating set.

Let nn be our desired power of x.x. Let sjs_j denote our current approximation of xn,x^n, with s0=gn.s_0 = g_n. We can find a better approximation sj+1s_{j+1} as follows.

Suppose ujku_j \in k is the leading coefficient of sjxns_j - x^n and let dj=deg(sjxn).d_j = \deg(s_j - x^n). If u=0,u = 0, then sj=xns_j = x^n and we’re done. Otherwise, take sj+1=sjujgdj.s_{j+1} = s_j - u_jg_{d_j}.

Note that each step of this procedure strictly decreases the degree of the error sjxn.s_j - x_n. Since a polynomial has a finite degree, this procedure terminates. This gives us

xn=gnj=1ujgdj,x^n = g_n - \sum_{j=1}u_jg_{d_j},

which expands to an k[t]k[t]-linear combination of elements of S.S. Any element of k[x]k[x] can, in turn, be produced by taking kk-linear combinations of the powers of x.x.

2.

Since kk is a field. it is Noetherian. By Hilbert’s basis theorem k[t]k[t] is Noetherian. Any subring RR of k[x]k[x] containing both ff and all kk-scalars is a k[t]k[t]-submodule of k[x]k[x] with k[t]k[t] acting via φ.\varphi. Since k[x]k[x] is a finitely generated module over a Noetherian ring, it is itself Noetherian (by III.6.8). Hence, RR is finitely generated.

3.

Let RR be a subring of k[x]k[x] containing k.k. Suppose RR contains no non-constant polynomials. Then RR is isomorphic to kk and hence Noetherian. Otherwise, it must contain at least one non-constant polynomial f.f. In such case, the arguments above show that RR is finitely generated as a k[t]k[t]-module.

Take an ideal II of R.R. As an k[t]k[t]-module, II is also finitely generated. But note that k[t]k[t] acts on RR by embedding k[t]k[t]-scalars in RR via φ.\varphi. We can just forget k[t]k[t] and use RR-scalars directly. Hence, II is finitely generated as an RR-module.

Thus, RR is Noetherian.

\blacksquare