OctoInformBeta

OpenAI preprint, September 2026 · §5. Extracting a derivation on the determinant ring

Lemma 5.1: On a domain over CC, nonzero products have additive degree and a function dividing its derivative is invariant

Derivation degree detects invariants.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Let RR be a commutative CC-algebra that is a domain, and let DD be a CC-linear derivation of RR. The issue is how repeated differentiation measures functions and constrains the factors of a function annihilated by DD.

A CC-linear derivation DD of RR is locally nilpotent if every element of RR is annihilated by some iterate, as in (1): ∀r∈R,∃n∈ℕ,Dn(r)=0.\forall r\in R,\qquad \exists n\in\mathbb N,\qquad D^n(r)=0.(1)✦ Here ℕ\mathbb N includes zero, and D0D^0 is the identity map.

For a locally nilpotent derivation DD of RR, the orbit polynomial of r∈Rr\in R is the polynomial ΦD(r;t)∈R[t]\Phi_D(r;t)\in R[t] defined by (2): ΦD(r;t)=∑n≥0Dn(r)n!tn.\Phi_D(r;t)=\sum_{n\geq 0}\frac{D^n(r)}{n!}\,t^n.(2) The derivation degree νD(r)\nu_D(r) of r∈Rr\in R is the degree in tt of its orbit polynomial, with value zero at r=0r=0, as specified in (3): νD(r)={deg⁡tΦD(r;t),r≠0,0,r=0.\nu_D(r)= \begin{cases} \deg_t\Phi_D(r;t),&r\ne0,\\ 0,&r=0. \end{cases}(3)✦

An invariant of DD is an element h∈Rh\in R with D(h)=0D(h)=0, and the set of invariants is denoted ker⁡D\ker D. A subring B⊆RB\subseteq R is factorially closed if, whenever nonzero f,g∈Rf,g\in R satisfy fg∈Bfg\in B, both ff and gg belong to BB.

The Claim

Theorem 1. Let RR be a commutative CC-algebra that is a domain, and let DD be a locally nilpotent CC-linear derivation of RR. For nonzero f,g∈Rf,g\in R, the product identity (4) holds: νD(fg)=νD(f)+νD(g).\nu_D(fg)=\nu_D(f)+\nu_D(g).(4)✦ For every h∈Rh\in R, the zero-degree criterion is (5): νD(h)=0⇔D(h)=0.\nu_D(h)=0\quad\Longleftrightarrow\quad D(h)=0.(5)✦ When D(h)≠0D(h)\ne0, the derivative satisfies (6): νD(D(h))+1=νD(h).\nu_D\bigl(D(h)\bigr)+1=\nu_D(h).(6)✦ For nonzero f,g∈Rf,g\in R with fg∈ker⁡Dfg\in\ker D, the conclusion is (7): D(f)=0andD(g)=0.D(f)=0\qquad\text{and}\qquad D(g)=0.(7)✦ For any h∈Rh\in R that divides D(h)D(h), the divisibility criterion (8) gives D(h)=0.D(h)=0.(8)✦

In the paper, the result reads: for a domain RR over CC and a locally nilpotent derivation DD, define νD(h)\nu_D(h) for nonzero hh as the largest nn with Dn(h)≠0D^n(h)\ne0; then νD(fg)=νD(f)+νD(g)\nu_D(fg)=\nu_D(f)+\nu_D(g) for nonzero f,gf,g, νD(D(h))=νD(h)−1\nu_D(D(h))=\nu_D(h)-1 when D(h)≠0D(h)\ne0, the kernel is factorially closed so that a nonzero invariant product has invariant factors, and D(h)≠0D(h)\ne0 implies D(h)∉hRD(h)\notin hR.

What Is Checked

The assertions concern locally nilpotent CC-linear derivations of commutative CC-algebras that are domains, with derivation degree defined from the orbit polynomial (2). For every r∈Rr\in R, the orbit coefficients have finite support, as stated in (9): {n∈ℕ:Dn(r)n!≠0}is finite.\left\{n\in\mathbb N:\frac{D^n(r)}{n!}\ne0\right\} \quad\text{is finite}.(9)✦