Closed

Closed submodules and controlled matrix images

Theorem 1.18.2. (Closed submodules and controlled matrix images). Let AA be a generalized KK-affinoid algebra. For every finite m≥0m\geq 0, every algebraic submodule N⊂AmN\subset A^m is finitely generated and closed. For every chosen finite generating list n1,…,ntn_1,\ldots,n_t of NN, there is a constant C>0C>0 such that ∀n∈N∃a1,…,at∈A:n=∑iaini,max⁡i∥ai∥≤C∥n∥.\forall n\in N\quad \exists a_1,\ldots,a_t\in A: \quad n=\sum_i a_in_i, \qquad \max_i\|a_i\|\leq C\|n\|. In particular, every finite matrix over AA has closed image and a uniform bound for actual lifts from the induced image norm. The constants may depend on the matrix and the chosen norms, but not on the element lifted. Moreover, AA is Noetherian.

Proof

Proof. We may suppose A≠0A\ne 0. Theorem 1.17.6 supplies a finite free tuple rr such that R=ArR=A_r, with its coefficient norm, is a strict norm-admissible KrK_r-affinoid algebra. This step uses the defining presentation of AA, not Noetherianity of AA. Theorem 1.11.5 therefore applies to RR with the indicated norm.

Let NRN_R be the algebraic RR-span of ιAm(N)\iota_A^m(N) in RmR^m. It is finite and closed by the strict finite-module theorem. Choose finitely many generators of NRN_R and expand each as a finite RR-linear combination of elements of NN. The union of these finite lists gives n1,…,nt∈Nn_1,\ldots,n_t\in N that generate NRN_R over RR. If n∈Nn\in N, write ιAm(n)=∑ibiιAm(ni)\iota_A^m(n)=\sum_i b_i\iota_A^m(n_i) and apply p0p_0 coordinatewise. Lemma 1.18.1 gives n=∑ip0A(bi)ni.n=\sum_i p_0^A(b_i)n_i. Thus NN is finite. The same argument applied to any constant vector lying in NRN_R proves N=(ιAm)−1(NR).N=(\iota_A^m)^{-1}(N_R). Since NRN_R is closed and ιAm\iota_A^m is continuous, NN is closed. Taking m=1m=1 proves Noetherianity.

Now let D:At→AmD:A^t\to A^m be any chosen generating matrix for NN. Its scalar extension Dr:Rt→RmD_r:R^t\to R^m has image NRN_R. By Theorem 1.11.5, there is C>0C>0 such that each ιAm(n)\iota_A^m(n) has a lift b∈Rtb\in R^t with ∥b∥≤C∥n∥\|b\|\leq C\|n\|. Put a=p0t(b)a=p_0^t(b). The coefficient lemma gives Da=nDa=n and ∥a∥≤∥b∥\|a\|\leq\|b\|. This is the asserted actual lifting bound. Finite matrices are bounded by the maximum of the norms of their entries; the reverse bound just proved is precisely norm-strictness onto the induced image. No algebraic flatness or completed-tensor exactness was used in the descent. ◻

Citation

Non-Archimedean Geometry Foundations, Tag NAG-001560, 1.18.2.

Proof in chapter · Fixed-edition record