gpt 6 astra ultra attempt on Math-479

Zilber’s Quasiminimality Conjecture for Complex Exponentiation

Definitions and mathematical statement

The complex exponential field is the first-order structure \[ {\mathbb{C}}_{\exp}=({\mathbb{C}};+ ,\cdot,0,1,\exp),\qquad \exp(z)=e^z. \] A subset $D\subseteq{\mathbb{C}}$ is definable with parameters if some first-order formula $\varphi(x,a_1,\ldots,a_m)$ in this language, with finitely many parameters from ${\mathbb{C}}$, satisfies $D=\{z:{\mathbb{C}}_{\exp}\models\varphi(z,a_1,\ldots,a_m)\}$. Quantifiers range over all complex numbers.

Quasiminimality asserts that every such one-variable definable set is countable or has countable complement. Countable includes finite. This is not the stronger claim that every definable set is finite or cofinite, nor a claim about arbitrary analytic subsets or arbitrary subsets of ${\mathbb{C}}$. Parameters are allowed, which is part of the requested scope.

Formulation and concept references: [S1].

Short English statement

Is every one-variable set definable using complex arithmetic and exponentiation either countable or the complement of a countable set?

Research attempt

Scope and source audit (27 September 2026).

The desired dichotomy concerns arbitrary first-order formulas, including alternations of quantifiers over all of $\mathbb C$, and permits any finite tuple of complex parameters. It is substantially stronger than the same statement for zeros of one exponential polynomial. The supplied Wilkie paper [S1] gives a conditional analytic-continuation route, not a completed proof for $\mathbb C_{\exp}$. In particular, its Theorem 1 assumes its Analytic Continuation Property. We do not assume that property below. The primary-source audit also distinguishes the quasiminimality theorem for complex powers [E1] from the exponential-field conjecture: the former uses a weaker structure and does not settle the latter.

This attempt proves the dichotomy for three explicit fragments, derives a precise common-exceptional-set reformulation of the full problem, and identifies where unrestricted existential projection defeats the elementary analytic argument. The arguments are elementary reconstructions, not claims of new special cases. No identification with a constructed pseudo-exponential field, and no transcendence conjecture, is assumed.

The countable--cocountable algebra.

Let $\mathcal A$ be the family of subsets of $\mathbb C$ that are countable or have countable complement. Countable means at most countable. This family is closed under complements and countable unions: if one member of a union is cocountable, so is the union; otherwise the union is countable. De Morgan's law gives closure under countable intersections. These assertions use the ordinary set-theoretic countable-union principle. They do not assert that every countable union is first-order definable.

A useful equivalent description is that any sequence $D_j\in\mathcal A$ has a countable exceptional set $B$ outside which the membership pattern in all the $D_j$ is constant. Indeed, take the union of $D_j$ when it is countable and of $\mathbb C\setminus D_j$ otherwise. Each summand is countable. There cannot be a set whose two sides are both countable, because $\mathbb C$ is uncountable.

A complete proof for quantifier-free formulas.

Proposition 475.1. Every subset of $\mathbb C$ defined by a quantifier-free formula in the stated language, with arbitrary finitely many complex parameters, belongs to $\mathcal A$.

Proof. Every one-variable term denotes an entire function. This follows by induction on terms: the variable and constants are entire, sums and products preserve entire functions, and so does composition with $\exp$. An atomic equality $t(x)=s(x)$ is the zero set of the entire function $f=t-s$. If $f$ is identically zero this is all of $\mathbb C$. Otherwise the identity theorem makes every zero isolated.

For completeness, an isolated subset of $\mathbb C$ is countable. Choose the countable basis of discs with rational real and imaginary parts of the centre and positive rational radii. For each isolated point, choose a basis disc meeting the subset only at that point. Assign the first such disc in an enumeration. Different points receive different discs, giving an injection into $\mathbb N$. Thus an atomic equality is countable or all of $\mathbb C$, and its negation is in $\mathcal A$. A quantifier-free formula is a finite Boolean combination of atomic equalities, so the preceding closure properties finish the proof. $\square$

The proof allows nested exponentials such as $\exp(\exp(x+a))$; no restriction on exponential depth was used. It is not an effective identity test for entire functions represented by terms. The mathematical alternative "identically zero or isolated zeros" does not provide an algorithm deciding the alternative.

The language printed above has no division symbol. If one works on the natural domain of rational expressions in such terms, denominators can be cleared while retaining explicit nonvanishing conditions. Each such condition is still a quantifier-free equality or inequality of entire terms. Omitting the domain conditions could introduce erroneous points.

Infinite definable sets are already necessary.

The set \[ H=\{z:\exp(z)=1\}=2\pi i\mathbb Z \] is countably infinite and has infinite complement. Thus strong minimality (finite or cofinite definable subsets) fails even in this elementary fragment. With the allowed parameter $\tau=2\pi i$, the formula $\exp(\tau x)=1$ defines precisely $\mathbb Z$ as a subset of $\mathbb C$. This does not contradict quasiminimality: a definable countably infinite set and its induced arithmetic are permitted by the proposed dichotomy.

Nor may one silently use the real line as a definable set. Both $\mathbb R$ and $\mathbb C\setminus\mathbb R$ are uncountable, so a definition of $\mathbb R$ in this structure, even with parameters, would disprove the target. Taking a real or imaginary part in a mathematical description is not automatically an operation in its first-order language. For the same reason the analytic condition $|x|<1$ cannot simply be added to a formula while claiming to remain in the original structure.

An unrestricted polynomial witness can be eliminated.

There is a fragment with a quantifier genuinely ranging over all of $\mathbb C$ that the same method handles completely. Let \[ P(x,y)=a_d(x)y^d+\cdots+a_1(x)y+a_0(x), \] where each $a_j$ is an entire term in $x$ and the fixed parameters. Proposition 475.2. The set defined by $\exists y\ P(x,y)=0$ belongs to $\mathcal A$. In fact it is defined by \[ \left(\bigvee_{j=1}^{d}a_j(x)\ne0\right) \quad\vee\quad \left(\bigwedge_{j=0}^{d}a_j(x)=0\right). \]

Proof. For each fixed $x$, if at least one coefficient with positive index is nonzero, $P(x,\cdot)$ is a nonconstant complex polynomial and has a root. If all those coefficients vanish, it has a root exactly when the constant coefficient vanishes, in which case every $y$ is a root. These are all possibilities. Proposition 475.1 applies to the displayed formula.$\square$

The proof does not apply to arbitrary systems by eliminating one equation at a time: a witness for one equation need not also solve another. Nor does it cover $\exp(y)$ occurring in the witness variable. As a simple transcendental instance that is nevertheless explicit, \[ \{x:\exists y\ \exp(y)=x\}=\mathbb C\setminus\{0\}. \] The exclusion of zero is immediate. For $x\ne0$, choose an argument of $x$ and take $y=\log|x|+i\arg x$. This is an existence proof, not a claim that a global definable choice of logarithm has been constructed.

Countable witness sets and a quantified fragment.

Proposition 475.3. Let $B_1,\ldots,B_m$ be fixed countable subsets of $\mathbb C$. Start with any quantifier-free formula $\theta(x,y_1,\ldots,y_m,a)$ and quantify its $y_j$ variables in any order using $\exists y_j\in B_j$ or $\forall y_j\in B_j$. The resulting set of $x$ belongs to $\mathcal A$.

Proof. For each fixed tuple of witnesses, Proposition 475.1 applies, since the witness entries become finitely many additional complex parameters. An existential quantifier over a countable set takes a countable union; a universal quantifier takes a countable intersection. Induction outward through the finite quantifier string proves the assertion. Empty witness sets cause no problem: the respective results are empty or all of $\mathbb C$.$\square$

If the $B_j$ are definable, these restricted quantifiers can of course be expressed by ordinary formulas. For example $H=\ker(\exp)$ is a parameter-free definable countable witness set. The proposition covers arbitrary finite alternations over that fixed kernel and allows arbitrary nested exponential terms in the matrix. It also covers fixed countable fibres $\{y:\exp(y)=c\}$, when parameters $c$ are fixed.

An important qualification is the word fixed. For the relation $\exp(y)=x$, each fibre over a nonzero $x$ is countable, but the union of all fibres is $\mathbb C$. The preceding proof does not turn an $x$-dependent countable family of witnesses into a fixed countable index set. A local choice of logarithm changes by $2\pi i\mathbb Z$, and the choice of that logarithm is precisely extra information.

One common exceptional set: an exact reformulation.

Let $A\subset\mathbb C$ be countable, and let $\operatorname{Form}_1(A)$ denote all formulas with one free variable and finitely many parameters from $A$. There are only countably many such formulas: the language is countable, every formula is finite, and finite tuples from $A$ form a countable set.

Proposition 475.4. The conjecture is equivalent to the following assertion for every countable $A$: there exists a countable set $C_A$ such that any two points outside $C_A$ satisfy exactly the same formulas in $\operatorname{Form}_1(A)$.

Proof. Assume quasiminimality. For each formula, take its countable side, either its realization set or the complement. Their countable union is $C_A$. Outside it, each formula has a constant truth value, so any two such points have the same full one-variable type over $A$.

Conversely, take a formula with a finite parameter set $A$ and a set $C_A$ as described. Its truth value is constant on the nonempty set $\mathbb C\setminus C_A$. The defined set is therefore contained in $C_A$, or its complement is contained in $C_A$. This is exactly the desired dichotomy.$\square$

This is a reformulation, not a construction of $C_A$. In particular it does not say that two points of the same type are interchanged by an automorphism of the given, nonsaturated structure. Such an automorphism statement would require its own homogeneity argument.

One can explicitly construct a countable field $F_A$ containing $A\cup\{2\pi i\}$ and closed under algebraic closure inside $\mathbb C$, exponentiation, and all logarithms of its nonzero elements. Start with the countable field generated by $A\cup\{2\pi i\}$, and repeatedly adjoin its algebraic elements, its exponential images, and the fibres of $\exp$ over its nonzero elements; then take the field generated by these elements. Each step remains countable: a countable field has countably many polynomials with finitely many roots each, and each nonempty exponential fibre is a coset of $2\pi i\mathbb Z$. The union of the stages has the claimed closure properties.

What is missing is a proof that all points outside this, or some larger countable set, have the same type. Closure under field operations, exponentiation, and logarithms does not control arbitrary quantified systems. Choosing $C_A=F_A$ without proving this extra assertion would simply conceal the main difficulty.

Where the analytic projection attack stops.

Suppose a system of entire equations $F(x,y)=0$, with $y\in\mathbb C^r$ and $F$ taking values in $\mathbb C^r$, has a solution $(x_0,y_0)$ with $\det(\partial F/\partial y)(x_0,y_0)\ne0$. The holomorphic implicit function theorem supplies an open disc $U$ around $x_0$ and a holomorphic witness $y(x)$ on $U$. Consequently its existential projection contains $U$. Any finitely many strict nonvanishing conditions true at $(x_0,y_0)$ remain true after shrinking $U$, by continuity.

This is genuine local information, but it is not cocountability. Even an open subset of $\mathbb C$ can have uncountable complement. The unit disc is a topological illustration of that logical gap, not an asserted exponential-field-definable counterexample. Singular witnesses present an additional problem, since the displayed Jacobian condition can fail.

Attempting to move the local witness along paths requires continuation and control of the exceptional set. A local logarithm demonstrates the distinction between local solvability and global single-valued choice. If a holomorphic $g$ on $\mathbb C^*$ satisfied $e^{g(z)}=z$, then differentiation would give $g'(z)=1/z$. Integrating around the unit circle yields $0=2\pi i$, impossible. Local logarithms can be continued along paths, but the resulting value depends on winding. General implicit exponential systems require substantially more than this well-understood example, and one cannot assume their witnesses extend along every desired path.

Wilkie's Analytic Continuation Property supplies a carefully qualified continuation statement along generic paths; his Theorem 1 then obtains quasiminimality using a back-and-forth argument. The elementary lemmas above neither establish that property nor replace the back-and-forth step. The source is used to locate the missing global mechanism, not to import its unproved hypothesis as a theorem about $\mathbb C_{\exp}$.

Adversarial checks and outcome.

The parameter issue was tested at each step: specializing witnesses creates only finitely many parameters per slice, whereas the common exceptional-set lemma explicitly takes the union over countably many formulas. The countable language alone cannot force quasiminimality: adding one unary predicate for $\mathbb R$ still gives a countable language and immediately gives a set with two uncountable sides. Similarly, countability of a generated subfield cannot imply uniformity of all types outside it.

What has been proved is the full dichotomy for quantifier-free formulas, for a single polynomial witness with entire-term coefficients, and for arbitrary finite strings of quantifiers over fixed countable witness sets. Proposition 475.4 gives an exact target for extending these results. The first unresolved step is uniform control of unrestricted complex witnesses and their continuation, strong enough to handle every formula over a fixed countable parameter set. No proof or counterexample for the full conjecture is supplied.

Sources

• S1 A. J. Wilkie, "Analytic continuation and Zilber's quasiminimality conjecture," arXiv:2306.14562 (26 June 2023). https://arxiv.org/pdf/2306.14562v1. Formulation source.
• E1 Francesco Gallinaro and Jonathan Kirby, Quasiminimality of complex powers, Forum of Mathematics, Sigma 12 $2024$, e123. https://arxiv.org/abs/2304.06450; https://doi.org/10.1017/fms.2024.82. Primary abstract and accepted-manuscript introduction inspected 27 September 2026. The powers structure is distinguished from the full complex exponential field.
• E2 Alex J. Wilkie, full-text HTML of the supplied source, https://arxiv.org/html/2306.14562v1. Definitions 1--3 and conditional Theorem 1 inspected 27 September 2026; the Analytic Continuation Property is not treated as proved for the structure needed here.