Alec Kriebel

An explicit complex polynomial retract in five variables: a consequence of affine-space cancellation failure

This note gives a polynomial map of complex affine five-space that becomes fixed after one application: applying it twice gives the same result as applying it once. Its image is a smooth complex fourfold whose coordinate ring is not a polynomial algebra in four variables.

The construction makes an inherited consequence explicit. OpenAI supplies the complex cancellation example, its one-variable stabilization and its nonpolynomiality. The classical cylinder-to-retract mechanism is credited to Nagamine. This note gives split maps and a factored five-component polynomial presentation.

Original abstract

OpenAI's explicit complex cancellation fourfold has a polynomial one-variable cylinder and is not affine four-space. We transport coefficient inclusion and evaluation at zero through its concrete cylinder coordinates. This gives a complex algebra retract of a five-variable polynomial ring, of transcendence degree four, that is not a polynomial algebra. We provide split maps and a five-component idempotent polynomial endomorphism, in factored form with no denominators, and exact symbolic certificates for the coordinate identities, including the fiber at the parameter zero. The implication from cancellation failure to a negative answer to Costa's retract question is classical; the nonpolynomiality input and cancellation breakthrough belong to OpenAI. This note makes that consequence explicit in ambient dimension five and makes no assertion about ambient dimension four.

A polynomial map that fixes its own image

A retraction sends the surrounding space onto a smaller space and fixes every point of that image. Here the five output coordinates are polynomials in five complex inputs. Once a point has reached the image, applying the same map again leaves it unchanged.

The image has complex dimension four and is smooth everywhere. Nevertheless, it is not complex affine four-space as an algebraic variety: its ring of polynomial functions is not ℂ[t₁,t₂,t₃,t₄]. Adding one free coordinate to the fourfold does give affine five-space. That cancellation failure is the input from OpenAI.

The word “retract” refers to the split polynomial maps and their idempotence. No linearity, orthogonality, distance minimization or nearest-point property is asserted.

The algebra and the precise result

Work over the full complex field ℂ. Set

P = ℂ[p,s,u,F,J],   x = s²+u³+p²F,

H = x²F−(1+2sx)J−p²J²−pu,   A = P/(H).

The theorem gives a smooth integral complex algebra A of transcendence degree four and explicit unital complex-algebra maps

i: A → B = ℂ[p,s,u,m,e],   r: B → A,   ri = idA.

The subalgebra i(A) is not a polynomial complex algebra. In particular, the idempotent algebra endomorphism ir has image exactly i(A). Its nonpolynomiality uses OpenAI's pinned cancellation theorem for this exact defining equation.

On affine schemes, the arrows reverse. The map π = Spec(i): ℂ⁵ → Spec(A) and the closed embedding σ = Spec(r): Spec(A) → ℂ⁵ satisfy πσ = idSpec(A). Thus ρ = σπ is an idempotent polynomial endomorphism of ℂ⁵. Its scheme-theoretic image is the closed copy σ(Spec(A)), with coordinate algebra B/ker(r) ≅ A.

This gives the negative instance of Costa's retract assertion over ℂ in five ambient variables. The note does not settle the corresponding four-variable question.

Why a cylinder gives a retract

The inherited cancellation theorem supplies a concrete isomorphism θ: A[w] → B even though A is not isomorphic to ℂ[t₁,t₂,t₃,t₄]. Coefficient inclusion ȷ: A → A[w] and evaluation ε at w = 0 obey εȷ = idA. Transporting them gives i = θȷ and r = εθ−1.

Therefore ri = idA and (ir)² = ir. This is the classical implication recorded in Nagamine's Proposition 1.4. The note's task is to write the transport as polynomial formulas for the specified example.

The factored five-component map

The following factors give the map without expanding large polynomials. For variables s,u,M,λ, put x0 = s²+u³ and

a0 = 4s²,   b0 = 1+2sx0,   c0 = 2sx0−1,   d0 = x0²,

fM(λ) = a0λ+b0M,   jM(λ) = c0λ+d0M,

CM = M²(2x0+6sx0²+4s²x0³−x0⁴),

q1,M = M[4x0a0b0−2s(a0d0+b0c0)−2c0d0],

q2 = 2x0a0²−2sa0c0−c0²,   Q1,M(λ) = q1,M+q2λ.

For inputs (p,s,u,m,e) use these factors with M = m and define

ℓ = pu−p²Cm+p³e,   f = fm(ℓ),   j = jm(ℓ),

W = −e−(u−pCm+p²e)Q1,m(ℓ)−pf³,

X = s²+u³+p²f,   D = −3u²W+3p²uW²−p⁴W³,

S = s+p²XD,   U = u−p²W,

𝓕 = f−(1+2sX)D−p²X²D²,   𝓙 = j−X²D.

For arbitrary (p,s,u,F,J), let

𝓜(s,u,F,J) = (1−2sx0)F+4s²J,   L = x0²F−(1+2sx0)J,

M = 𝓜(s,u,F,J),   Q = 2x0F²−2sFJ−J²,

𝓔(p,s,u,F,J) = −pF³−Q1,M(L)(u−pQ−p³F³).

The five output coordinates are

ρ(p,s,u,m,e) = (p,S,U, 𝓜(S,U,𝓕,𝓙), 𝓔(p,S,U,𝓕,𝓙)).

In each helper evaluation, x0 and its factors are recomputed from the helper's own s,u arguments. All expressions are polynomials; no factor introduces division by p, m or any other coordinate. The notation above makes the manuscript's helper dependence on M explicit.

The map is global, including at p = 0

The cylinder proof uses polynomial identities to avoid a formula defined only after inverting p. In the zero fiber, let κ be the bracket in q1,m; it is a polynomial, not a quotient requiring m ≠ 0. Put W = −e−uκm. Then

ρ(0,s,u,m,e) = (0,s,u, m+3u²W, −uκ(m+3u²W)).

The resulting point has its new W equal to zero, so the second application fixes every component. The global identity ρ² = ρ follows from the split maps, and this formula also displays the limiting fiber directly.

Smoothness holds on p ≠ 0 by the Laurent polynomial coordinates. At p = 0 the F,J partial derivatives of H are x0² and −(1+2sx0). They cannot both vanish because 4s²x0²+(1−2sx0)(1+2sx0) = 1.

Attribution and limits

The result concerns this explicit complex fourfold and its embedding in complex affine five-space. It supplies no real-form, positive-characteristic or ambient-four-variable theorem.

The nonpolynomiality, defining example and stabilization are inherited from OpenAI. The general cylinder-to-retract implication is classical. This note's contribution is the explicit transport, factored map and checkable presentation; no independent cancellation breakthrough or firstness claim is made.

The verification package contains symbolic checks of coordinate identities and finite rational regression examples. Those examples do not prove nonpolynomiality or replace the polynomial identities and split-map proof. The written theorem is the basis of the consequence. An independent upstream Lean kernel build was not reproduced, and formal verification of these retract maps is not claimed.

Original files and manuscript

  1. Source and verification package (ZIP, 153,906 bytes) · Zenodo original ZIP
  2. Manuscript (PDF, 74,424 bytes) · Zenodo original PDF

Open the six-page PDF · Download PDF · Zenodo record

These unchanged originals are listed in the record's order. The manuscript and declared publication date are 6 October 2026; record metadata reports creation on 7 October 2026 UTC and updates on 10 October 2026. The hosting page displays edition v1; no author-supplied version number is stated in the record metadata.

Credited sources

The manuscript contains the broader bibliography. The attribution of this particular example and of the classical mechanism is separate from any historical-priority assessment.

Citation and rights

Kriebel, Alec. (2026). An explicit complex polynomial retract in five variables: a consequence of affine-space cancellation failure [Unrefereed preprint; Zenodo display edition v1]. Zenodo. doi:10.5281/zenodo.23203270. Download BibTeX.

The original note and original verification package are licensed CC BY 4.0. The bundled OpenAI Lean sources and adapted configuration/audit harness have the explicit Apache License 2.0 exception described in LICENSES.md and THIRD_PARTY_NOTICES.md. Cited publications retain their own rights.

Status and collaboration

AI tools were used extensively in research, drafting, symbolic verification and adversarial audits. This is an unrefereed, AI-assisted preprint; automated reviews are not conventional human peer review. The author is responsible for the claims, and the upstream work is credited to OpenAI.

Independent verification of the polynomial transport and scheme-image description, substantive improvements to the exposition, and development of a journal version would be useful contributions. Collaboration and authorship would be based on actual contributions.