Zariski's Multiplicity Conjecture

Zariski's Multiplicity Conjecture

OpenAI's claim

Manuscript 1 (family 059, 2026-09-27): Two isolated hypersurface germs in four complex variables are claimed to have the same ambient topological type while their multiplicities are four and five. Manuscript 2 (family 059, 2026-09-24): Two isolated weighted-homogeneous hypersurface germs in a common finite dimension divisible by eight are claimed to be ambiently homeomorphic despite having multiplicities two and three.

Mathematical statement of the claimed result

Mathematical claim 1.

OpenAI claims reduced convergent holomorphic germs $f_4,f_5:(\mathbb C^4,0)\to(\mathbb C,0)$ for which zero is an isolated critical point in each case, such that \[ \operatorname{ord}_0f_4=4,\qquad\operatorname{ord}_0f_5=5,\qquad (\mathbb C^4,V(f_4),0)\cong(\mathbb C^4,V(f_5),0) \] by a homeomorphism germ of ambient pairs. The multiplicity $\operatorname{ord}_0f$ is the least total degree appearing among nonvanishing terms of its power series. The equivalence is embedded topology, despite distinct multiplicities.

Mathematical claim 2.

For some finite integer $N>3$ divisible by eight, OpenAI claims reduced real-coefficient polynomials $f_1,f_2\in\mathbb R[z_1,\ldots,z_N]$, each weighted homogeneous for positive rational weights and with an isolated critical point at zero, satisfying \[ \operatorname{ord}_0f_1=2,\qquad\operatorname{ord}_0f_2=3,\qquad (\mathbb C^N,V(f_1),0)\cong(\mathbb C^N,V(f_2),0) \] by an ambient-pair homeomorphism germ. Weighted homogeneous means $f(t^{a_1}z_1,\ldots,t^{a_N}z_N)=t^df(z)$ for positive integers $a_i,d$. The existence assertion does not specify a fixed small ambient dimension.

Outline of the argument

Manuscript 1 (family 059, 2026-09-27): Convergent equations in analytic charts give a finite-map lattice calculation that matches the integral Seifert forms. The nonspherical open-book classification on the seven-sphere identifies the embedded links, and the ambient cone construction transfers that identification to the zero-set germs.

Manuscript 2 (family 059, 2026-09-24): A finite integer relation builds chain-polynomial seeds with matching spectra but different lowest degrees. Tensor operations and local--global isometries align their integral Seifert forms, with determinant cancellation enabling strong approximation; polynomial sums realize the operations, and knot classification extends the resulting link equivalence to the germs.

Formal verification

Family 059 has no Lean scope page in this snapshot; no formalization of these manuscript results is listed.

This site has not executed the Lean files or checked their dependencies. These descriptions report the scope of the linked declarations, rather than a new verification of the proofs.

Status caveat

OpenAI attributes these results to an internal model. Its release notice cautions that results lacking full formalization can contain mistakes. This site has not independently assessed the proofs; a solved label records the scope of the reported claim.

Sources

• OpenAI, Ambiently homeomorphic isolated hypersurface germs in $\mathbb C$${}^{4}$ with multiplicities four and five, 2026-09-27; citation key OAI:Ambiently-homeomorphic-isolated-hypersurface-germs-in-C4-with-multiplicities-four-and-five-September-27-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: thm:main.
• OpenAI, Ambiently homeomorphic isolated hypersurfaces of multiplicities two and three, 2026-09-24; citation key OAI:Ambiently-homeomorphic-isolated-hypersurfaces-of-multiplicities-two-and-three-September-24-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: thm:main.
• OpenAI catalogue entry.

Zariski's Multiplicity Conjecture

Definitions and mathematical statement

Let $f,g:({\mathbb{C}}^n,0)\to({\mathbb{C}},0)$ be nonzero reduced holomorphic germs. Their multiplicities are their orders of vanishing, namely the least total degrees of nonzero terms in their Taylor expansions. Embedded topological equivalence means there is a germ of a homeomorphism of ambient neighborhoods, fixing zero, that maps $V(f)$ onto $V(g)$. The conjecture is \[ ({\mathbb{C}}^n,V(f),0)\cong_{\mathrm{top}}({\mathbb{C}}^n,V(g),0) \quad\Longrightarrow\quad\operatorname{mult}_0f=\operatorname{mult}_0g. \] Reducedness is essential because taking a power preserves the zero set while changing the order of vanishing. No isolated-singularity or quasihomogeneity hypothesis is included in the full target.

Formulation and concept references: [S1].

Short English statement

Does the embedded topological shape of a reduced complex hypersurface singularity determine its multiplicity?

Research attempt: multiplicity from slices and torus-knot data

GPT-6 Astra Ultra, 27 September 2026. Status: UNRESOLVED.

Multiplicity is easy to recover from analytic coordinates and from certain special links. The general question is whether an arbitrary ambient homeomorphism retains enough of that information. I prove analytic invariance, work out a complete reconstruction for coprime binomial plane branches using their Alexander polynomial, and explain why neither the Milnor number alone nor a generic-slice argument extends automatically to arbitrary reduced hypersurfaces.

Analytic coordinate changes preserve the leading degree.

Write $f=f_m+f_{m+1}+\cdots$ with $f_m$ nonzero homogeneous of degree $m$. If $\phi$ is a biholomorphic coordinate germ with derivative $A\in\operatorname{GL}_n(\mathbb C)$ and $u$ is a holomorphic unit, then \[ u\,(f\circ\phi)=u(0)f_m(Az)+\text{terms of degree greater than }m. \] The leading term is nonzero because $A$ is invertible and $u(0)\ne0$. Thus multiplication by a unit and analytic coordinate change preserve multiplicity. For reduced hypersurfaces, analytic equivalence of the zero sets supplies precisely such a unit relation between reduced defining equations.

This proof cannot be applied to a homeomorphism germ by replacing $A$ with a derivative: a homeomorphism need not have a derivative at the origin, much less an invertible complex-linear one. The conjecture asks for a topological invariant replacing this leading-term calculation, not a formal Taylor expansion of an arbitrary coordinate homeomorphism.

Reducedness also has an immediate adversarial test. The germs $f$ and $f^r$, for $r>1$, have the same zero set, while $\operatorname{ord}_0(f^r)=r\operatorname{ord}_0(f)$. Thus even identity of the embedded zero sets cannot recover multiplicity without the reducedness assumption. No argument involving the set alone can repair this failure in the nonreduced setting.

Generic complex lines read the multiplicity exactly.

Choose a vector $v$ with $f_m(v)\ne0$. Then \[ f(tv)=t^m f_m(v)+O(t^{m+1}), \] so the restriction to the complex line in direction $v$ has order $m$. Such directions form the complement of the projective zero locus of $f_m$ and are therefore generic. Equivalently, the local intersection number of the hypersurface with a generic complex line is $m$.

This supplies a clear conditional route. If the ambient equivalence were known to identify suitable generic complex slices together with their local intersection multiplicities, the desired equality would follow. But an arbitrary homeomorphism sends a complex line to a topological disc, not necessarily a complex analytic line or curve transverse to the tangent cone. It does not identify the projective exceptional set $\{f_m=0\}$ of bad directions. Counting intersections along an arbitrary image disc is not the same analytic operation as taking the order of $g(tv)$.

Nor does a single nongeneric slice determine multiplicity. For $f(x,y)=y^2-x^3$, a generic line has order two, while the line $y=0$ gives order three. The same singularity already exhibits both answers depending on the direction. This finite example rules out a proof that chooses a slice after a homeomorphism without checking its required genericity.

A complete binomial plane-branch subcase.

Let \[ f_{p,q}(x,y)=x^p-y^q,\qquad p,q>1,\quad\gcd(p,q)=1. \] The branch is reduced and irreducible, parametrized by $x=t^q$, $y=t^p$, and its multiplicity is $\min(p,q)$. Its link in a small three-sphere is the torus knot of type $(p,q)$. The classical link and Alexander-polynomial facts used here are recorded in Milnor's treatment [E1]. Its Alexander polynomial, normalized with constant term one, is \[ \Delta_{p,q}(T)= \frac{(T^{pq}-1)(T-1)}{(T^p-1)(T^q-1)}. \tag{1} \] One can check polynomiality directly through cyclotomic factors. Coprimality implies that the common factor of $T^p-1$ and $T^q-1$ is $T-1$. After cancellation, precisely the factors $\Phi_d(T)$ with $d\mid pq$, $d\nmid p$ and $d\nmid q$ remain. Thus the right side is a monic integral polynomial.

Here is an explicit reconstruction of the multiplicity from $1$, rather than just an appeal to the torus-knot classification. Every root of $\Delta_{p,q}$ has order dividing $pq$, and primitive $pq$th roots occur because $p,q>1$. Therefore the largest order of a root is \[ P=pq. \] The breadth of the polynomial, invariant under multiplication by a Laurent monomial, is \[ b=pq+1-p-q=(p-1)(q-1). \] Hence $p+q=P+1-b$, and the unordered pair $\{p,q\}$ is recovered as the two roots of \[ U^2-(P+1-b)U+P=0. \tag{2} \] In particular $\min(p,q)$ is determined.

Embedded topological equivalence of two such plane branches preserves their local knot Alexander invariant, up to the conventional units and reversal of the variable. These operations preserve both root orders and breadth. Applying $2$ proves equality of multiplicities in this entire binomial subcase. The proof uses both invariants of the polynomial; its degree alone would not suffice.

Equal Milnor numbers do not give equal multiplicities.

The Milnor algebra of $x^p-y^q$ is \[ \mathbb C\{x,y\}/(x^{p-1},y^{q-1}), \] with basis $x^iy^j$ for $0\le i\le p-2$, $0\le j\le q-2$. Its dimension is $\mu=(p-1)(q-1)$. Thus \[ x^2-y^7\quad\text{and}\quad x^3-y^4 \] both have Milnor number six, but their multiplicities are two and three. Their Alexander polynomials are respectively \[ T^6-T^5+T^4-T^3+T^2-T+1, \qquad T^6-T^5+T^3-T+1. \tag{3} \] The largest root orders are fourteen and twelve, so the reconstruction above distinguishes them. These germs are not embedded topologically equivalent; they are a countertest to the weaker proposed criterion based only on $\mu$.

This distinction also prevents a misuse of results about $\mu$-constant families. A theorem about all fibers of a specified family cannot be applied to an arbitrary pair merely because the two Milnor numbers agree: the required family with its hypotheses must first be supplied. Similarly, a theorem for isolated singularities cannot be used when the critical locus has positive dimension. The full question deliberately includes that latter case.

A further boundary case with smooth branches.

For a reduced plane-curve germ whose $r$ irreducible branches are all smooth, the multiplicity is $r$. Indeed each reduced branch equation has order one, and the product equation has order equal to the sum of the orders, since the product of the nonzero leading homogeneous terms is nonzero. The number of branches is the number of local connected components away from the singular point, or of link components, and is topologically invariant. Consequently the conjecture also holds for a pair of plane-curve germs both having only smooth branches, regardless of their tangencies.

The qualifier about both germs is intentional in this elementary proof. The component count by itself does not tell whether a branch is smooth: a cusp and a smooth branch each have one component. Knot information or another theorem is needed to recognize that distinction topologically. In ambient dimension one, reducedness forces a germ vanishing at zero to have order one, giving another immediate boundary case.

Scope of the recent source.

The inherited source [S1], in its November 2025 version, proves results for specified quasihomogeneous parametrized hypersurfaces. Its Theorem 1.1 assumes irreducible quasihomogeneous surfaces in $\mathbb C^3$, smooth normalization, and only transverse double points away from the origin. Its higher-dimensional theorem adds finite determinacy, corank-one and multiple-point-space conditions. Those are substantial hypotheses absent from the universal statement above. The paper does not claim that every reduced hypersurface admits such a parametrization.

The calculations here likewise concern restricted classes. A topological invariant of a link may give rich information in dimension two without automatically determining the leading homogeneous degree in every dimension. The generic-line criterion remains analytic, while the Alexander reconstruction exploits the special two-exponent formula $1$. Neither construction supplies a general multiplicity functional on all embedded topological hypersurface germs.

Verification and first unresolved step.

The exact script divides the polynomials in $1$, checks $3$, records their cyclotomic root orders, and reconstructs $(p,q)$ for a finite range of coprime exponents. It also counts the monomial bases of the Milnor algebras and verifies the nongeneric cusp slice. These are checks of the proved formulas, not a search over topological equivalences.

The completed work is analytic invariance, the generic-line characterization, a topological reconstruction for binomial plane branches, and the all-smooth-branch subcase. The first unresolved step toward the full problem is a topological replacement for the generic complex-slice condition, or an invariant that recovers multiplicity without that condition for arbitrary reduced hypersurfaces. No such invariant is constructed here. The universal Zariski multiplicity conjecture remains unresolved by this attempt.

Sources

• S1 Zariski's multiplicity conjecture for quasihomogeneous hypersurfaces with non-isolated singularities. Accessed 2026-09-01. https://arxiv.org/html/2309.02624. Formulation source.
• E1 J. Milnor, Singular Points of Complex Hypersurfaces, Annals of Mathematics Studies 61, Princeton University Press, 1968. Used for the local link, torus-knot and Alexander-polynomial framework; the reconstruction from root orders and breadth is given explicitly above. https://doi.org/10.1515/9781400881819.