Flat local maps cannot lower multiplicity

← All digests

Family 194 · Algebra
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-21

1. The question

A singular point has an algebraic thickness: its tiny neighborhoods contain more information than equally small neighborhoods of a smooth point. Hilbert–Samuel multiplicity measures the leading coefficient of that local growth.

Lech’s conjecture asks whether this thickness can decrease under a flat local homomorphism. Flatness is an algebraic condition expressing the absence of certain new relations when coefficients are extended. It is the natural language for many well-behaved families, but it does not directly compare the two maximal-ideal filtrations used to measure multiplicity.

Family 194 claims that multiplicity never decreases under such a map, in every dimension and characteristic. The approach works with complexes and normalized lengths rather than requiring special finite modules over every source ring.

2. Background and essential definitions

Measuring infinitesimal neighborhoods

Let $(A,\mathfrak a)$ be a nonzero Noetherian local ring of dimension $d$. The quotient $A/\mathfrak a^N$ describes the point together with information up to order $N-1$. Its finite length $\ell_A(A/\mathfrak a^N)$ counts residue-field composition factors.

For large $N$, this length grows like a polynomial of degree $d$. The Hilbert–Samuel multiplicity is

$$e(A)=\lim_{N\to\infty} \frac{d!\,\ell_A(A/\mathfrak a^N)}{N^d}.$$

It is a positive integer. For the regular power-series ring $k[[x_1,\ldots,x_d]]$, the leading growth is $N^d/d!$, so the multiplicity is one. In dimension zero, multiplicity is simply the length of the ring.

Multiplicity is a numerical measure of local thickness, not a complete classification of singularities. Distinct local rings can have the same multiplicity and very different geometry.

Local maps and flatness

A homomorphism $(R,\mathfrak m)\to(S,\mathfrak n)$ is local when the maximal ideal of the target pulls back to that of the source. It is flat when tensoring an exact sequence of $R$-modules with $S$ preserves exactness. This prevents the coefficient extension from creating unexpected linear relations.

The closed fiber is $S/\mathfrak mS$, the part lying over the source’s distinguished point. A flat local map may change dimension and residue field; the conjecture does not require the map to be a finite extension.

The difficulty is that flatness naturally controls $S/\mathfrak m^NS$, whereas the definition of $e(S)$ uses $S/\mathfrak n^N$. Since $\mathfrak mS$ and $\mathfrak n$ can differ, comparing lengths from flatness is not already the desired theorem.

Why modules and complexes enter

A maximal Cohen–Macaulay module has depth equal to the ring’s dimension: a full sequence of parameters acts without creating zero divisors successively. An Ulrich module is such a finite nonzero module whose multiplicity equals its minimal number of generators. That equality can turn a geometric growth question into a generator-count comparison.

A finite free complex is a sequence of finite-rank free modules linked by matrices whose consecutive products are zero. Its homology measures the failure of exactness. Complexes can retain useful numerical information even when no single finite module has all the desired properties.

3. Development history and previous partial results

Lech introduced the multiplicity question in work from 1960 and 1964. His results established the inequality for sources of dimension at most two and in important situations with complete-intersection closed fiber. Standard reductions connect his equal-dimensional formulation with the more general flat-local statement.

Ma proved the conjecture for equal-characteristic sources of dimension three. In arbitrary equal-characteristic dimension $d$, he also obtained the weaker bound

$$e(R)\le\max\{1,d!/2^d\}\,e(S).$$

This controls multiplicity but loses the sharp coefficient one in higher dimensions. His use of Cohen factorizations and Hilbert–Kunz multiplicity is a predecessor of the characteristic-$p$ argument in the new manuscript.

Hanes developed Ulrich-module methods and asymptotic versions in which generator counts approach multiplicities. Ma’s weakly lim Ulrich sequences relax exact Cohen–Macaulay behavior to suitable negligible homological errors. Their existence over a local domain implies Lech’s inequality for all flat local targets.

Ma used these methods to prove the conjecture when the source is a standard graded algebra over a perfect field, localized at its homogeneous maximal ideal. The target need not be graded. Meng obtained a complementary graded-target result under specified homogeneity and equal-dimension assumptions.

Passing to associated graded rings does not make these theorems universal: the induced graded map need not remain flat. Moreover, Yhee constructed complete two-dimensional nonnormal domains without weakly lim Ulrich sequences. Thus the existence of such sequences over every source is not available as a general solution.

Iyengar, Ma, and Walker developed inequalities for short free complexes using lim Ulrich methods. Their construction combines a finite free resolution with a parameter Koszul complex, providing the immediate model for the new approach. The manuscript also cites the finite-module counterexample in family 195 as a further reason to distinguish module existence from complex-based positivity; that counterexample is not an assumption proving Lech’s inequality.

4. What the new manuscripts claim

The manuscript, dated 23 September 2026, claims that every flat local homomorphism of nonzero Noetherian local rings satisfies

$$e(R)\le e(S).$$

There is no restriction on dimension, residue fields, or characteristics. The source and target need not be domains, complete, graded, or Cohen–Macaulay in the final statement. Those properties appear only in reductions or auxiliary constructions.

The argument’s main technical claim is a uniform lower bound for certain finite free complexes over an auxiliary ring equipped with an additive normalized length. Its constants depend on the number of parameters and free ranks, rather than on the auxiliary extension or the particular length theory. Additional homological and stability hypotheses on that length are required; additivity alone is not enough.

A related positive-characteristic consequence

For a short free complex $F$ over a complete local domain of characteristic $p>0$, the manuscript also claims

$$\chi_\infty(F)\ge e(R).$$

Here a short complex is concentrated in degrees $0,\ldots,d$, has finite-length homology, and has nonzero zeroth homology. Its Dutta multiplicity $\chi_\infty$ is the limit of its Euler characteristic after raising differential entries to $p^n$-th powers and dividing by $p^{nd}$.

This consequence is the complete-domain positive-characteristic case of a conjecture of Iyengar, Ma, and Walker. The manuscript does not claim that entire short-complex conjecture for every complete local ring. Nor does it use the full conjecture as an assumption for the main theorem.

5. Why the advance matters

Provisional importance: 5/5. If correct, this resolves a central comparison problem in local algebra and replaces dimension-dependent losses with the sharp universal coefficient one.

The theorem would establish a robust monotonicity principle for local multiplicity under flat maps. Such maps occur throughout algebraic geometry and commutative algebra, so a statement without characteristic or dimension restrictions has a broad range of potential applications.

The methodological significance is the move from ideal finite modules to numerical estimates on complexes. That route can remain meaningful even when a desired Cohen–Macaulay or Ulrich module does not exist. It also links characteristic-$p$, mixed-characteristic, and characteristic-zero techniques within one comparison framework.

6. The main idea

Make a complex carry the multiplicity comparison

After reductions, the proof works with a complete domain source and equal source and target dimensions. A Cohen factorization introduces an auxiliary complete local ring $T$ and an ideal $J$ with $S=T/J$, while keeping $e(T)=e(R)$.

A finite free resolution $P$ of $S$ is tensored with a Koszul complex on parameters of $S$. The parameters are chosen so that their ideal has the same multiplicity as the target maximal ideal. The combined complex has the appropriate length, degree-zero rank one, and homology supported at the closed point.

This single complex is where the source-side lower estimate and target-side upper estimate can meet.

Obtain a bound independent of the auxiliary ring

The core estimate considers differential matrices whose entries lie deep in a parameter ideal $I$, say in $I^s$. The auxiliary length $\lambda$ makes higher parameter Koszul homology negligible and gives controlled growth of parameter quotients. In simplified notation, the claimed bound has the shape

$$\lambda(H_0F)\ge \lambda(C/I)s^h \left(b_0-\frac{A_h\sum_i b_i}{s}\right),$$

where $h$ is the number of parameters, $b_i$ are free ranks, and $A_h$ depends only on $h$. The error disappears as the differential entries become deeper.

The manuscript obtains this through a blowup, which separates powers of the parameter ideal geometrically, and a vector bundle with prescribed cohomological vanishing. Its construction builds on Eisenbud and Schreyer’s work. The chosen cohomology pattern makes an Euler characteristic nonnegative, while carefully balanced leading coefficients leave the desired contribution of the zeroth free rank.

The auxiliary ring may be non-Noetherian. Accordingly, the comparison uses finite filtrations and stabilized images, and tracks when negligible-length objects remain negligible under sums and limits. These technical conditions protect the numerical estimate during passage to large extensions.

Three characteristic regimes

In characteristic $p$, Frobenius raises matrix entries to arbitrarily high powers without changing free ranks. Normalized lengths on a perfection, obtained by adjoining compatible $p$-power roots, relate the resulting zeroth homology to Hilbert–Kunz multiplicity. Componentwise inequalities and multiplicity identities then yield the target comparison.

Mixed characteristic has no Frobenius endomorphism of the same kind. The manuscript instead uses roots of parameters, integral-extension results of Heitmann, and perfectoid normalized lengths. Auxiliary extension ranks can grow, so lengths are divided by those ranks before comparison.

The order of limits matters. For a fixed extension, parameter powers first tend to infinity to obtain a target-side upper bound. Only afterward does the depth of the differential entries increase. The upper bound stays independent of the chosen extension, allowing the sharp coefficient to survive.

In residue characteristic zero, the argument reduces to a finite free map over a common coefficient field. A finite system of polynomial identities records algebra structure, freeness, locality, and bounds in the required directions on the two multiplicities. Artin approximation and specialization transfer a hypothetical strict counterexample to positive characteristic. That would contradict the already established characteristic-$p$ case.

7. Scope, limitations, and open questions

Flatness and locality are essential hypotheses. A quotient map or an arbitrary homomorphism need not preserve multiplicity in the claimed direction. The theorem also does not assert strict increase or classify all equality cases.

The multiplicities use each ring’s own maximal ideal. Replacing that by the extended source ideal changes the quantity and misses the central issue. Likewise, an argument on associated graded rings must justify flatness there rather than assume it passes automatically.

The proof does not claim that every local domain admits an Ulrich module, a small Cohen–Macaulay module, or a weakly lim Ulrich sequence. Its auxiliary lengths can vanish on nonzero modules, and the associated algebras need not be finite or Noetherian.

The short-complex consequence has the stated complete-domain and positive-characteristic scope. Broader short-complex inequalities, effective descriptions of equality in Lech’s inequality, and sharper bounds using fiber information remain separate problems.

8. Sources and further reading

Lech’s Note on multiplicities of ideals (1960) and Inequalities related to certain couples of local rings (1964) formulate the classical comparison. Ma’s dimension-three work and lim Ulrich approach give major earlier cases and methods.

Iyengar, Ma, and Walker’s short-complex inequalities are a direct predecessor. Yhee’s nonexistence examples explain a limitation of universal module-sequence approaches.

Eisenbud and Schreyer’s vector-bundle and Betti-table work and Cai, Lee, Ma, Schwede, and Tucker’s perfectoid length work supply important inputs. The pinned source below states the uniform complex estimate and the complete claimed deduction of Lech’s inequality.

Source manuscripts

• M1. OpenAI, Lech’s multiplicity conjecture (September 23, 2026).

Sources consulted on 21 October 2026 at repository revision 1581d4c. Download the manuscript citations.