Why a proper algebraic intersection should have positive multiplicity

← All digests

Family 193 · Algebra
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-21

1. The question

Two transverse curves in a plane meet with multiplicity one. If they are tangent, the meeting counts more heavily. Algebraic geometry extends this local counting idea to higher dimensions and to arithmetic spaces where ordinary pictures no longer suffice.

Serre’s definition uses an alternating sum of algebraic correction terms. The first term is positive, but later terms can subtract from it. Why should a properly dimensioned, isolated intersection still have a strictly positive total?

Family 193 claims a positive answer over every regular local ring, including the general ramified mixed-characteristic case. The theorem removes additional lifting and geometric assumptions from previously known cases. It concerns the strict inequality needed to distinguish a genuine intersection from a total that cancels to zero.

2. Background and essential definitions

A ring describing one neighborhood

A local ring $(R,\mathfrak m)$ has one maximal ideal $\mathfrak m$, representing a distinguished point. Its residue field is $R/\mathfrak m$. A Noetherian ring has every ideal finitely generated, making its algebraic geometry locally manageable.

The Krull dimension $d$ is the maximum length of a chain of prime ideals. A local ring is regular when its maximal ideal can be generated by exactly $d$ elements. This is the algebraic analogue of a nonsingular ambient space.

A finitely generated $R$-module can describe a geometric piece with multiplicities. For instance, $R/I$ describes the equations in an ideal $I$. Its support is the collection of primes where it remains nonzero after localization, and its dimension is the dimension of that support.

A module of finite length has a finite composition series; its length counts the simple factors. Over a local ring those factors are copies of the residue field. Geometrically, finite length means that the module is concentrated at the distinguished point, with a finite algebraic size.

Correcting the naive intersection

The tensor product $M\otimes_R N$ is a first approximation to intersecting two modules. Tensoring does not always preserve exact sequences. The groups $\operatorname{Tor}_i^R(M,N)$ measure this failure: take a free resolution of one module, tensor it with the other, and compute the resulting homology.

If $M\otimes_R N$ has finite length and $R$ is regular of dimension $d$, Serre’s intersection multiplicity is

$$\chi_R(M,N)=\sum_{i=0}^{d}(-1)^i \operatorname{length}_R\operatorname{Tor}_i^R(M,N).$$

All the lengths are finite, and higher terms vanish. The degree-zero term is the tensor product itself. The alternating signs are why positivity is a theorem to seek, rather than part of the definition.

For a concrete example, take $R=k[[x,y]]$, the formal power-series ring in two variables. The curves $y=0$ and $y=x^r$, with $r\ge1$, give a quotient intersection $k[[x]]/(x^r)$, of length $r$. Here the higher Tor correction vanishes, so the multiplicity is $r$. The general problem allows much more complicated modules and arithmetic ambient rings.

Mixed characteristic

In equal characteristic, a local ring and its residue field have the same characteristic. In mixed characteristic, the ring has characteristic zero and the residue field characteristic $p>0$. For a regular local ring, the ramified case has $p\in\mathfrak m^2$. Arithmetic and local geometry then interact in ways not visible over a single coefficient field.

3. Development history and previous partial results

Serre introduced the Tor Euler characteristic and proved positivity in equicharacteristic and unramified settings using reduction to the diagonal. His method also treats formal power-series rings over arbitrary complete discrete valuation rings. Thus “ramified” does not mean that every example in that class was previously unknown; the remaining issue is the unrestricted mixed-characteristic situation.

Roberts, and independently Gillet and Soulé, proved the vanishing theorem when the dimensions of the two modules sum to less than the ambient dimension. Their methods use local intersection-theoretic invariants and Adams operations. This result removes lower-dimensional terms in later reductions.

Gabber proved nonnegativity in the remaining mixed-characteristic case using alterations and an intersection calculation on a normal bundle. Nonnegativity is a major advance, but it still permits zero for complementary-dimensional modules. Strict positivity must exclude that last possibility.

Further partial results impose geometric or lifting hypotheses. Skalit proved positivity for regular local rings essentially smooth over a two-dimensional regular local base. KC and Soto Levins proved positivity under a Serre-liftability condition: the module comes from an unramified ambient ring with the appropriate codimension relationship. Their condition is weaker than requiring the usual Tor-vanishing lift, but it is still an additional hypothesis.

Another approach seeks sequences of finite modules that become approximately Cohen–Macaulay after normalization. Cohen–Macaulay behavior means that a full sequence of parameters has no unwanted homological obstruction. Bhatt, Hochster, and Ma showed that appropriate sequences for both quotient domains imply positivity. Perfectoid towers provide related constructions in specified settings.

The present manuscript instead works with complexes after discarding sequences whose normalized free ranks become negligible. It credits normalized-length methods to Faltings and subsequent almost-ring and perfectoid work, and uses algebraic K-theory descent to compare numerical invariants.

The introduction also records a December 2025 public assertion of the prime-quotient statement as a same-scope claim, rather than treating it as an established predecessor theorem. The digest follows that attribution without assessing competing proofs or priority.

4. What the new manuscripts claim

The manuscript, dated 23 September 2026, claims the following unrestricted positivity statement. Let $(R,\mathfrak m)$ be a commutative Noetherian regular local ring of dimension $d$, and let $M,N$ be nonzero finitely generated modules. If

$$\operatorname{length}_R(M\otimes_RN)<\infty, \qquad \dim_R M+\dim_R N=d,$$

then

$$\chi_R(M,N)>0.$$

There is no characteristic, residue-field, or ramification restriction, and neither module must lift to an unramified regular local ring.

Two stated consequences

The manuscript applies a theorem of Kurano and Roberts to symbolic powers. For a ramified mixed-characteristic regular local ring and prime ideals $P,Q$ with $\sqrt{P+Q}=\mathfrak m$ and $\operatorname{ht}P+\operatorname{ht}Q=d$, it obtains

$$P\cap Q^{(r)}\subseteq\mathfrak m^{r+1}\quad(r\ge1), \qquad Q^{(r)}=Q^rR_Q\cap R.$$

Here localization at $Q$ ignores behavior away from that prime; the symbolic power measures order of vanishing at its generic point. The containment is a consequence of the positivity input, not a separate ingredient establishing it.

A second consequence compares intersection multiplicity with the product of Hilbert–Samuel multiplicities, which measure leading local growth. For prime quotients $A/P,A/Q$ meeting in complementary dimensions, set

$$G=\operatorname{gr}(A/P)\otimes_{\operatorname{gr}A}\operatorname{gr}(A/Q).$$

The associated graded ring records successive orders in the maximal ideal, geometrically the tangent cone. Under the additional hypothesis $\dim G\le1$, the manuscript claims

$$\chi_A(A/P,A/Q)\ge e(A/P)e(A/Q),$$

with equality exactly when $\dim G=0$. The dimension restriction is essential to the stated corollary; this is not an unrestricted product lower bound.

5. Why the advance matters

Provisional importance: 5/5. If correct, the theorem completes a central positivity problem in local algebra and arithmetic intersection theory. It would make the intuitive positive count of a proper intersection valid across all regular local ambient rings.

Its consequences connect an Euler characteristic to ordinary ideal containments and tangent-cone geometry. This illustrates why positivity matters beyond a single numerical inequality: once zero cancellation is excluded, established geometric identities yield stronger algebraic conclusions.

The method is also designed to avoid demanding a finite Cohen–Macaulay module on every singular quotient. Related finite-module existence questions have their own obstructions, discussed in family 195. An asymptotic construction can retain the numerical features needed for intersection theory without producing such a module.

6. The main idea

Replace ordinary size by normalized size

Standard reductions reduce the essential case to two complete local domains $D=R/P$ and $E=R/Q$, whose positive dimensions sum to $d$. Their intersection is supported at the closed point.

The manuscript constructs large auxiliary algebras with additive nonnegative normalized lengths. One can think of these lengths as module sizes divided by the increasing ranks of a tower of extensions. A nonzero module may have normalized length zero, so vanishing in this numerical sense is weaker than actual vanishing.

The useful property is that quotients by parameter sequences have positive finite normalized length, while their higher Koszul homology has length zero. A Koszul complex is the finite complex recording the compatibility of multiplication by the chosen parameters. These algebras behave like Cohen–Macaulay objects for the numerical calculation, even when actual finite modules with that property are unavailable.

Compare averages of conjugate evaluations

Finite extensions with Galois group actions allow the proof to compare ordinary intersection multiplicity with normalized Tor Euler characteristics. Algebraic K-theory organizes the additive relations between complexes needed for this comparison.

The comparison concerns sums over conjugates. It does not assert that the normalized length is invariant under every Galois action, nor that each individual evaluation equals the original multiplicity. After the first comparison, it remains to prove that each term in the relevant sum is positive.

For one such term, choose parameters from the second prime that cut the first domain to finite normalized length. Approximate the resulting module by finite-length modules $M_n$, dividing their ordinary sizes by numbers $s_n$. Their higher resolution ranks become negligible compared with $s_n$.

The proof then works in a category that identifies free sequences of rank $o(s_n)$ with zero. Truncating the resolutions produces supported complexes in that quotient. Support can emerge only after passing to the quotient, so the numerical evaluations must be built there, including for direct summands.

Obtain strict positivity rather than a nonnegative limit

A second comparison, now using the auxiliary algebra for $E$, eliminates higher homological contributions in normalized length. The remaining degree-zero contribution needs a positive lower bound.

All approximating modules are killed by one fixed power $\mathfrak m^a$. If $\beta_0(M_n)$ is their minimal generator count, then

$$\operatorname{length}(M_n) \le\beta_0(M_n)\operatorname{length}(R/\mathfrak m^a),$$

with the corresponding quotient local ring used in the actual argument. A positive limiting normalized length therefore forces a positive normalized generator count. That prevents the final nonnegative quantity from shrinking to zero.

The two conjugate-sum comparisons transfer this strict lower bound back to the original intersection multiplicity. The role of the common annihilator is concrete: it converts positivity of total normalized size into positivity of the surviving degree-zero term.

7. Scope, limitations, and open questions

Regularity of the ambient ring, nonzero finite modules, finite-length tensor product, and complementary dimensions are all part of the theorem. Dropping them changes the problem; arbitrary singular ambient rings need not support the same finite Tor formula or positivity behavior.

The theorem gives strict positivity, hence a positive integer, rather than an unrestricted optimal lower bound in terms of the separate module multiplicities. The product bound in section 4 retains its finite tangent-cone-support restriction.

The auxiliary normalized lengths may vanish on nonzero objects. Their use does not assert that the large algebras are ordinary finite Cohen–Macaulay modules, or that normalized exactness is literal exactness. Likewise, the Galois averaging argument must not be shortened to invariance of individual evaluations.

Further quantitative intersection bounds and broader equality criteria remain different questions from the central positivity claim explained here.

8. Sources and further reading

Serre’s Local Algebra, Chapter V develops the definition and classical cases. Roberts’ vanishing theorem and Gillet–Soulé’s Adams-operation approach establish the dimension-deficient result.

Skalit’s two-dimensional-base theorem and KC–Soto Levins’ lifting theorem provide earlier positivity cases. Kurano and Roberts’ symbolic-power theorem supplies the first consequence above.

Bhatt, Hochster, and Ma’s lim Cohen–Macaulay manuscript and Cai, Lee, Ma, Schwede, and Tucker’s perfectoid length work explain the asymptotic background. The present manuscript states its own normalized-length construction and comparison arguments, with those literature inputs attributed separately.

Source manuscripts

• M1. OpenAI, Positivity of Serre’s Intersection Multiplicity (September 23, 2026).

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