Height-one membership in a free target
Let \(R\) be a Noetherian unique factorization domain, let \(E\) be torsion-free, and let \(M\subseteq E\) be free. Assume full generic span in the scalar form \[ \forall x\in E,\quad\exists c\ne0,\quad cx\in M. \] Then \[ x\in M\quad\Longleftrightarrow\quad x/1\in M_{\mathfrak p} \text{ for every prime ideal }\mathfrak p\text{ of height one}. \] Equivalently, \(M\) is the intersection of the preimages of those localized submodules in \(E\).
Relation to the OpenAI notes
General module interface for the local membership test.
OpenAI notes: The free-target lemma uses a polynomial coefficient ring and a finite free ambient module. Basis coordinates in the fraction field lie in every height-one local ring, so their denominators are units. edge:lattice.
Lean (primary): The free-target interface allows a torsion-free ambient module over a UFD and states generic span through nonzero scalar multipliers. Its proof removes prime factors from those multipliers. A separate projective-target version obtains finite-support coordinates through a free-module section and retraction. ModuleAlgebra:45, ModuleAlgebra:106, IntegralPairings:1904.
Role in the proof: The proofs below expose the exact generic-span and prime-denominator witnesses used by local-to-global membership.
Proof
Step 1. Localize an existing member
If \(x\in M\), the fraction \(x/1\) has a numerator in \(M\), so it belongs to every \(M_{\mathfrak p}\).
Step 2. Obtain scalar witnesses at the principal prime ideals
Assume all the localized membership conditions. Let \(p\) be any prime element. Its ideal \((p)\) is a prime of height one.
The hypothesis at \((p)\), followed by the denominator-clearing calculation, gives \(a\notin(p)\) with \(ax\in M\). Equivalently \(p\nmid a\).
Step 3. Apply prime-denominator membership
The generic-span assumption gives a nonzero \(c\) with \(cx\in M\). Step 2 gives the required witness for every prime element.
The prime-denominator theorem applies to the free target \(M\), which is projective, inside the torsion-free ambient module \(E\). It yields \(x\in M\).
Step 4. Express the result as an intersection
A vector lies in the preimage of \(M_{\mathfrak p}\) exactly when its localized image \(x/1\) lies in that submodule. Membership in the intersection means that this holds for every height-one prime. The equivalence just proved identifies the intersection with \(M\).
If \(R\) is a field, every nonzero generic denominator is a unit. The generic-span condition then gives \(M=E\), agreeing with the intersection over the empty family of height-one primes.
For \(R=k[X,Y]\) and \(M=XR\subseteq R\), the prime \((X)\) tests divisibility by \(X\): \(ax\in XR\) with \(X\nmid a\) implies \(X\mid x\). At a different height-one prime, \(X\) is a unit and the localized ideal fills the localized ring.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.EdgeLattice.exists_denominator_of_localized_mem: ModuleAlgebra.lean:89.
OAI.KLInvariance.EdgeLattice.mem_iff_localized_at_height_one: ModuleAlgebra.lean:106.
OAI.KLInvariance.EdgeLattice.eq_iInf_height_one_localizations: ModuleAlgebra.lean:129.