Family 064 — Constant Milnor number and an unchanged surface singularity

← All digests

Family 064 · Algebraic and complex geometry
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-12

1. The question

Suppose the equation of a singular surface varies holomorphically with a parameter. A nearby smooth level of the equation has a finite amount of middle-dimensional topology, measured by the Milnor number. If that number stays constant, must the singularity remain topologically unchanged?

For isolated hypersurfaces, a classical theorem answers this outside complex surface dimension. The surface case lies exactly where the standard high-dimensional topological argument cannot provide the required product structure. Family 064 claims that missing case.

The conclusion is stronger than saying every individual zero set is homeomorphic to the first one. The manuscript claims a continuously varying family of ambient homeomorphisms, with continuously varying inverses, preserving both the parameter and the numerical value of the defining function. It therefore identifies the zero sets and all nearby levels together.

2. Background and essential definitions

The local family

A holomorphic germ records a function near a point, allowing its domain to be shrunk. Consider a holomorphic function $F(x,t)$, where $x=(x_1,x_2,x_3)\in\mathbb C^3$ and $t$ lies in a small complex disk $\Delta$. Write

$$f_t(x)=F(x,t),\qquad X_t=\{x:f_t(x)=0\}.$$

Three ambient complex variables give a hypersurface of complex dimension two, hence the term surface singularity. Assume $f_t(0)=0$ and the origin is an isolated critical point of every $f_t$.

The Milnor number is the finite integer

$$\mu(f_t,0)=\dim_{\mathbb C} \frac{\mathbb C\{x_1,x_2,x_3\}} {(\partial f_t/\partial x_1,\partial f_t/\partial x_2,\partial f_t/\partial x_3)}.$$

The denominator is the ideal generated by the spatial derivatives. Finiteness expresses isolatedness. Topologically, the Milnor fiber, obtained by taking a small nonzero level near the origin, has the homotopy type of a bouquet of $\mu$ two-spheres.

For example, $f=x_1^2+x_2^2+x_3^2$ has derivative ideal $(x_1,x_2,x_3)$, so $\mu=1$. The invariant is easy to define but does not obviously remember how the whole singular family sits in its ambient space.

Several meanings of unchanged

Topological right equivalence of two functions means one becomes the other after a homeomorphism of the source, without changing the target value. A family is topologically right-trivial if these homeomorphisms can be selected jointly continuously, together with their inverses.

This is a local topological statement. It does not require holomorphic coordinate changes, differentiability at the singularity, or distance control. Individual homeomorphisms for each parameter would not suffice: they could fail to fit together continuously.

A resolution replaces a singular surface by a smooth one, mapping isomorphically away from the singular point. Its exceptional curves record where the point has been replaced. A simultaneous resolution carries out this process throughout a family, sometimes after a finite change of parameter such as $t=u^e$.

3. Development history and previous partial results

Lê–Ramanujam proved that constant Milnor number gives constant embedded topological type when the hypersurface dimension is not two. Their comparison of Milnor fibers produces a homologically trivial cobordism. In high dimensions the h-cobordism theorem makes it a product; curves can be handled by lower-dimensional topology. Complex surfaces remain outside that argument. Timourian and King established the corresponding local triviality of function families outside the exceptional dimension. M1 cites the original papers for both stages.

A stronger numerical hypothesis was already known to imply geometric regularity. Let $\mu^{(j)}$ denote the Milnor number of a restriction to a general $j$-dimensional linear subspace. For a three-variable germ, $\mu^{(3)}=\mu$, while $\mu^{(1)}$ is multiplicity minus one. Teissier and Briançon–Speder related constancy of the full sequence to Whitney regularity, a condition controlling limiting tangent spaces and secant lines. Their examples also show that topological triviality need not imply Whitney regularity. Briançon–Speder.

Parusiński proved topological triviality for constant-Milnor-number families of the special form $f(x)+tg(x)$ in every dimension. Arbitrary holomorphic parameter dependence is broader than this linear form. Parusiński.

Fernández de Bobadilla–Pełka proved that finite constant Milnor number forces constant multiplicity, even for much more general continuous formal families. For surfaces this supplies constancy of $\mu^{(1)}$, but not automatically of the plane-section invariant $\mu^{(2)}$. Their theorem is a major ingredient in the history and in M1’s treatment of certain singularities, rather than a complete solution of the topological family problem. Equimultiplicity theorem.

Another route uses resolution invariants. Wahl introduced a logarithmic characteristic number for normal surface singularities. Okuma proved that its constancy gives simultaneous resolutions in the non-log-canonical Gorenstein case after finite base change. M1’s main numerical step is to derive constancy of that invariant from $\mu$ alone. Its introduction also credits recent semistable-degeneration results of Aldasoro Rosales and Peters’s symplectic result under a no-vanishing-folds hypothesis. Peters.

4. What the new manuscripts claim

Let $F$ be any holomorphic one-parameter family as above, with an isolated critical point at the origin for every parameter and

$$\mu(f_t,0)=\mu(f_0,0).$$

M1 claims that, after shrinking the disk and representatives, there is a homeomorphism germ

$$\Phi(x,t)=(\phi_t(x),t)$$

such that

$$\phi_t(0)=0,\qquad \phi_0=\operatorname{id},\qquad f_t(\phi_t(x))=f_0(x).$$

Both $\Phi$ and $\Phi^{-1}$ are jointly continuous on neighborhoods of the origin section over the original parameter disk.

Several details belong to the strength of the claim. The parameter dependence can be arbitrary and holomorphic; it need not be linear. The singularities need not be log canonical. The map is ambient in $\mathbb C^3$, so it controls how each surface sits in its neighborhood. It preserves the defining function, so no parameter-dependent change of the target value is inserted.

Although the proof uses finite covers of the parameter disk to obtain resolutions, the final result is asserted over the original disk. Triviality only after replacing $t$ by $u^e$ would be a weaker conclusion because monodromy could obstruct descent.

This theorem is local near the distinguished singular section. It does not assert a global trivialization of polynomial fibers throughout $\mathbb C^3$ or control behavior at infinity.

5. Why the advance matters

Importance: 5/5. If correct, the claim would complete the classical isolated-hypersurface $\mu$-constant picture in the dimension excluded by the Lê–Ramanujam argument, with the stronger function-preserving family conclusion.

It would show that one integer, constant under deformation, controls much more than the homotopy rank of nearby smooth fibers. The family would admit a coherent ambient topological identification, including its singular point and all nearby function levels.

The result would also connect three kinds of information: a local derivative algebra defining $\mu$, logarithmic intersection theory on resolutions, and continuous ambient transport. That bridge is particularly significant in complex dimension two, where direct high-dimensional cobordism methods are unavailable.

It does not say that equal Milnor numbers classify arbitrary singularities. The existence of the holomorphic constant-$\mu$ family is a substantive hypothesis.

6. The main idea

Convert a constant integer into a constant resolution invariant

For a non-log-canonical surface singularity, the source studies the canonical divisor plus the full reduced exceptional boundary on a resolution. A relative Zariski decomposition separates a nef part from an effective exceptional part. Its local nef cycle $P_{\mathrm{loc}}$ determines Wahl’s invariant

$$\beta=-P_{\mathrm{loc}}^2.$$

“Nef” here means nonnegative intersections with the curves relevant to the relative problem. The square is computed using the exceptional intersection matrix. The manuscript recalls resolution independence in the convention needed for Okuma’s theorem.

To compare this number across the family, it reduces the scalar calculation to algebraic curves of polynomial equations and compactifies them to projective surfaces without adding unwanted singularities. These individual finite-determinacy comparisons concern numerical invariants; they are not already a continuous trivialization of the original family.

Positive contributions must all vanish

Semistable reduction and the threefold log minimal model program produce a central fiber with one component dominating the original surface and additional projective surfaces $S_a$ over the singular point. Each has a carefully retained reduced boundary $B_a$.

After cancellation of the main component, constant Milnor number yields an Euler-characteristic relation

$$0=\sum_a\chi_c(S_a\setminus B_a)+\delta, \qquad\delta\ge0.$$

The compactly supported Euler characteristic $\chi_c$ is the version suited to noncompact boundary complements. The correction $\delta$ comes from interior smoothings. A logarithmic surface inequality gives

$$0\le(K_{S_a}+B_a)^2 \le3\chi_c(S_a\setminus B_a).$$

Thus every term contributing to the extra logarithmic squares must vanish. Specialization of intersection numbers then makes $\beta$ constant. Keeping the complete boundary is essential because both the Euler cancellation and the inequality must refer to the same open surfaces.

From resolved surfaces to ambient functions

Okuma’s theorem supplies simultaneous resolutions after finite base change. The manuscript shows that the exceptional curves form the required relative normal-crossing configuration, then constructs transport on resolved neighborhoods. This descends to punctured zero fibers and extends to nearby nonzero levels while preserving the function value.

Properness and reverse transport control both directions at the singular point. An additional correction removes the angular mismatch introduced by $t=u^e$, returning the trivialization to the original disk.

The log-canonical and double-point cases are handled separately. Equimultiplicity, spectral constancy, and sectional Milnor-number arguments supply the relevant regularity; splitting off a square reduces the double-point case to plane curves. The source does not force every singularity through the non-log-canonical resolution theorem.

7. Scope, limitations, and open questions

The hypotheses require a holomorphic one-complex-parameter family, three spatial variables, an isolated critical point along the fixed origin section, and finite constant Milnor number. The theorem does not address arbitrary nonisolated singularities or multiparameter families with their additional compatibility questions.

A continuous gradient vector field alone would not suffice: continuous fields can lack unique flows. The manuscript explicitly separates smallness near the singular section from the trajectory and reverse-transport control needed for a homeomorphism. The stated conclusion includes both maps and inverses.

Topological right-triviality does not imply Whitney regularity, analytic equivalence, or a bi-Lipschitz trivialization. In particular, the generic plane-section Milnor number is not assumed constant in the general case.

The claim is compatible with family 059, which concerns isolated pairs of equivalent germs with different multiplicities. Such a pair cannot be joined by a continuous constant-$\mu$ family under the established equimultiplicity theorem. A path hypothesis and a pairwise comparison are different problems.

Questions beyond this statement include nonisolated settings, parameter spaces of higher dimension, and stronger regularity of the trivializing maps under additional assumptions.

8. Sources and further reading

M1’s Theorem 1.1 gives the function-preserving trivialization. Sections 3–7 derive constancy of the logarithmic resolution invariant, Sections 8–9 construct and descend transport, and Section 10 supplies the remaining singularity cases. Appendix B explains why gradient estimates need additional flow control.

The primary links in Section 3 cover Whitney regularity, linear-parameter triviality, equimultiplicity, and the recent symplectic partial result. M1’s bibliography also gives the original Lê–Ramanujam, Timourian, King, Wahl, and Okuma papers that establish the classical problem and the resolution bridge.

Source manuscripts

• M1. OpenAI, Topological triviality of mu-constant families of surface singularities (September 24, 2026).

Sources consulted on 12 October 2026 at repository revision fd4aeeb. Download the manuscript citations.