Generic derivations, differential largeness, and NTP$_2$

We compare Fornasiero and Terzo’s framework of generic derivations on algebraically bounded structures with León Sánchez and Tressl’s differentially large fields. We show in the case of a single derivation that genericity and differential largeness coincide for éz-fields, as introduced by Walsberg and Ye. We also show that an NTP$_2$ algebraically bounded structure remains NTP$_2$ after expanding by a generic derivation.

💡 Research Summary

The paper investigates the relationship between two recent notions in the model theory of differential fields: generic derivations on algebraically bounded structures (as introduced by Fornasiero and Terzo) and differentially large fields (as defined by León Sánchez and Tressl). The authors focus primarily on the case of a single derivation and show that, for a special class of fields called “éz‑fields” (large fields in which every definable set is a finite union of étale‑open subsets, a notion introduced by Walsberg and Ye), the two concepts are equivalent.

The first part of the paper clarifies the definitions. A derivation δ on a field K is called generic if for every L(K)‑definable set X ⊆ K^{1+r} whose projection onto the first r coordinates has full dimension r, there exists a ∈ K with (a,δa,…,δ^{r}a) ∈ X. Differential largeness requires (1) that the underlying field K be large (i.e., every infinite K‑point set carries a non‑discrete étale‑open topology) and (2) that any field extension L of K in which K is existentially closed remains existentially closed as a differential field when equipped with the same derivation. The paper points out that these two notions can diverge when the language L contains extra symbols beyond the pure ring language, because genericity quantifies over L‑definable sets while differential largeness only sees L_ring‑definable sets.

To bridge this gap, the authors employ the étale‑open topology. Proposition 2.3 gives a topological characterization of differential largeness: for every smooth irreducible K‑variety V and every étale‑open subset X of its prolongation τV, if the projection of X has non‑empty interior in the étale‑open topology on V(K), then there is a K‑rational point a ∈ V(K) with (a,δa) ∈ X. This condition is shown to be equivalent to the original definition of differential largeness.

Using this characterization, the authors prove two corollaries. Corollary 2.4 shows that if K is large and δ is generic, then (K,δ) is differentially large. Corollary 2.5 proves the converse for éz‑fields: if K is an éz‑field and (K,δ) is differentially large, then δ must be generic. Thus, on éz‑fields the two notions coincide. This result extends earlier observations that held only under additional model‑completeness assumptions and applies, for example, to perfect Frobenius fields, which are known to be éz‑fields.

The second major contribution concerns the preservation of certain “neostability” properties under the addition of a generic derivation. Fact 1.4 lists several properties (model completeness, quantifier elimination, stability, NIP, distality, simplicity, rosiness, NSOP₁) that are known to transfer from an algebraically bounded theory T to its expansion T_δ^g (the theory of T together with a generic derivation). However, the transfer of tree‑type properties such as TP₂ (the tree property of the second kind) and NA‑TP (absence of the antichain tree property) had not been established in full generality.

Theorem 3.1 proves that if the expanded theory T_δ^g has TP₂, then the original theory T already has TP₂. The proof proceeds by translating an L_δ‑formula witnessing TP₂ into an L‑formula of the form ψ(∇^r x,∇^s y), where ∇^r x denotes the tuple (x,δx,…,δ^r x). By choosing the minimal r for which such a formula has TP₂, the authors analyze an L_δ‑indiscernible array and use algebraic boundedness to show that the projection of the associated definable sets must have full dimension. A Noetherian argument on Zariski closures then yields a contradiction unless r = 0, which would make ψ an L‑formula already having TP₂. Hence TP₂ cannot be introduced by adding a generic derivation.

Theorem 3.2 establishes an analogous result for NA‑TP, showing that the antichain tree property also transfers from T_δ^g to T. Consequently, any algebraically bounded theory that is NTP₂ (or NA‑TP) remains so after expanding by a generic derivation.

In summary, the paper makes two substantial advances. First, it identifies a natural class of fields (éz‑fields) on which generic derivations and differential largeness are equivalent, thereby clarifying the precise relationship between the two frameworks. Second, it demonstrates that adding a generic derivation to an algebraically bounded structure preserves sophisticated model‑theoretic complexity measures such as NTP₂ and NA‑TP. These results deepen the connection between differential algebra and model theory, providing a robust toolkit for studying differential expansions of tame fields and opening avenues for further exploration of stability‑theoretic phenomena in differential contexts.