Counting modular representations from local weights

← All digests

Family 202 · Algebra
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-21

1. The question

A finite group describes a collection of symmetries. A representation turns those abstract symmetries into matrices, making linear algebra available. Over the complex numbers, every representation splits into irreducible pieces. Over a field of prime characteristic, that comfortable picture can fail: a representation may contain an invariant subspace without having an invariant complementary subspace.

The irreducible pieces still exist, but their organization becomes much more subtle. Can their number be recovered from smaller groups associated with the prime? Alperin’s weight conjecture predicts that it can. Its blockwise version makes a separate prediction for each natural component, or block, of the representation theory.

Family 202 contains one manuscript, dated 23 September 2026. It claims the numerical blockwise conjecture for every finite group and every prime. The surprising feature of its approach is that a problem about counting representations is translated into a divisibility problem for covers of algebraic curves.

2. Background and essential definitions

Representations in characteristic p

Fix a finite group $G$, a prime $p$, and an algebraically closed field $k$ of characteristic $p$. The group algebra $kG$ consists of formal sums $\sum_{g\in G}a_g g$, with multiplication extended from the group law. A representation of $G$ over $k$ is the same thing as a module over $kG$.

A nonzero module is simple if its only submodules are itself and zero. The simple modules are the irreducible representations being counted. Their characters in modular representation theory are called irreducible Brauer characters. For the counting question, one can think of a Brauer character as a label for a simple module; its precise evaluation uses group elements whose orders are prime to $p$.

A block is an indecomposable direct factor of the algebra. More explicitly, a primitive central idempotent $e$, satisfying $e^2=e$, determines the factor $kGe$. Centrality ensures that every simple module belongs to exactly one block. Write $l(B)$ for the number of simple modules in block $B$.

The local objects called weights

A $p$-subgroup $Q$ has order a power of $p$; the trivial subgroup is allowed. Its normalizer is

$$N_G(Q)=\{g\in G:gQg^{-1}=Q\}.$$

The quotient $N_G(Q)/Q$ records symmetries preserving $Q$, after symmetries already inside $Q$ have been removed.

For a positive integer $a$, let $a_p$ denote the largest power of $p$ dividing it. A weight is a pair $(Q,\varphi)$, where $\varphi$ is an ordinary irreducible character of $N_G(Q)/Q$ satisfying

$$\varphi(1)_p=|N_G(Q)/Q|_p.$$

Here an ordinary character comes from characteristic-zero representation theory, and $\varphi(1)$ is the representation’s dimension. This equality is the defect-zero condition. It singles out local representations whose dimensions absorb the entire $p$-part of the quotient’s order.

Conjugating both the subgroup and its character produces an equivalent weight. We therefore count simultaneous $G$-conjugacy classes, not every subgroup embedding separately. A standard procedure, Brauer block induction, assigns each weight to a block of $G$. Let $W_p(B)$ denote the resulting set of classes assigned to $B$.

This assignment is important: simply adding up weights of normalizers without tracking their blocks would give a weaker, unrefined counting question.

3. Development history and previous partial results

The local viewpoint grows out of Richard Brauer’s modular representation theory. A block has an associated defect group, a $p$-subgroup measuring how far its structure is from the particularly simple defect-zero case. Brauer correspondence relates a block to a distinguished block in the normalizer of its defect group. These constructions suggested that substantial information about a large group should already be visible in its local subgroups.

Alperin formulated the weight conjecture at the 1986 Arcata conference; the account appeared in 1987. It extended a notion of weights arising for finite groups of Lie type to arbitrary finite groups. Alperin and Fong subsequently established major families of cases, including symmetric groups and finite general linear groups in odd modular characteristic.

Knörr and Robinson supplied a different organizing language in 1989: alternating sums over chains of $p$-subgroups. This resembles inclusion–exclusion. Counting contributions with alternating signs can remove overlapping local information and expose exactly the discrepancy that the conjecture says should vanish.

Later reduction theorems shifted much of the work to finite simple groups. Navarro and Tiep developed a reduction for the non-blockwise conjecture; Späth obtained a blockwise reduction. Such reductions do not themselves prove the conjecture: the relevant conditions for the simple groups, including compatibility with automorphisms, still have to be established.

The manuscript cites subsequent progress for special linear and special unitary groups, groups with abelian Sylow $3$-subgroups, certain exceptional groups, and inertial blocks. These results illustrate both the breadth of the evidence and the technical work needed to pass from one family to another.

The proposed approach takes a different route. It uses subgroup-chain cancellation but then moves to geometry, rather than completing a classification-based list of inductive conditions. Its geometric ingredients have their own history: Frobenius’s character formulas count group homomorphisms from surface groups; period–index theory constructs curves with restricted divisor degrees; Tate’s genus-change theory controls inseparable maps in positive characteristic. The new manuscript claims to combine suitably strengthened versions of these tools into a uniform divisibility statement.

4. What the new manuscripts claim

The central assertion is

$$l(B)=|W_p(B)|$$

for every prime $p$, every finite group $G$, and every $p$-block $B$. There are no stated restrictions on the defect groups or the type of finite group. Weights with $Q=1$ are included.

This is an equality of cardinalities. It does not assert a preferred bijection between the two sets, and it does not require such a bijection to respect every automorphism or Galois action. These stronger compatibility questions should be distinguished from the numerical theorem.

The paper also supplies consequences previously known to follow from the conjecture. If $D$ is a defect group of $B$, and $b$ is its Brauer correspondent in $N_G(D)$, then

$$l(B)\geq l(b),\qquad D\text{ abelian}\ \Longrightarrow\ l(B)=l(b).$$

For a Sylow $p$-subgroup $P$, it obtains a corresponding inequality between the total numbers of irreducible Brauer characters of $G$ and $N_G(P)$. Some special cases of these consequences were already theorems; their significance here is the removal of the remaining restrictions through the claimed general result.

A final functorial identity packages alternating local contributions more richly. The manuscript credits an existing theorem of Boltje, Bouc and Yılmaz for reducing that identity to the numerical conjecture. It is therefore presented as a consequence using earlier machinery, rather than as an independently invented reformulation.

5. Why the advance matters

Importance: 5/5, provisional. The score reflects a claimed resolution of a central local–global conjecture in modular representation theory, with no restriction on finite groups or primes.

An equality of counts may look modest beside a classification of all representations. Yet such equalities constrain the possible structures of blocks and reveal which local objects contain the right amount of information. They provide tests and inputs for more refined conjectures, even when they do not construct the representations individually.

The method would also be notable. A numerical statement about finite-dimensional algebras is linked to arithmetic properties of curves over carefully chosen fields. The value lies both in the headline conjecture and in a proposed geometric mechanism capable of forcing exact integer identities.

This importance assessment concerns mathematical significance if the claims hold. It does not measure how explicit the resulting correspondence is: the theorem is numerical, and the geometric proof is not offered as a practical enumeration algorithm.

6. The main idea

Turn a discrepancy into tuple counts

The algebraic reduction starts with the difference between the desired local and global counts. Alternating sums over subgroup chains cancel repeated contributions. Calculations in group algebras then express the remaining discrepancy through traces of powers of central elements.

Expanding a trace can turn an algebraic expression into a count of tuples. The tuples here lie in a finite group $J$ with no nontrivial normal $p$-subgroup. Write this condition as $O_p(J)=1$. For $n=p^s$, the tuples have the form $(x,y,u_1,\ldots,u_n)$ and satisfy

$$[x,y]u_1\cdots u_n=1,\qquad \langle x,y,u_1,\ldots,u_n\rangle=J.$$

The commutator is $[x,y]=xyx^{-1}y^{-1}$. Each $u_i$ has order divisible by $p$. Tuples are identified by simultaneous inner conjugation. Their positions remain ordered, even when a multiset of conjugacy classes is prescribed.

The target is strong divisibility: for each fixed $h$, all the relevant counts should be divisible by $p^h$ once $s$ is sufficiently large. Crucially, the threshold must work uniformly over the prescribed multisets. Otherwise an increasing number of uncontrolled terms could spoil the trace argument.

Interpret tuples as covers

The displayed relation is the fundamental-group relation for a punctured surface of genus one. The generating condition makes the associated cover connected. Thus the tuples can be interpreted as monodromy descriptions of connected covers, with group $J$, of a punctured genus-one curve.

The paper constructs a curve in characteristic $p$ whose divisor degrees are all divisible by the large power $p^s$. A divisor is a finite integer combination of points, with degrees accounting for their fields of definition. This divisibility is called a large index.

At the same time, the marked points vary independently in the relevant field-differential directions. These are differentials of the field parameters, not many independent regular differentials on a single genus-one curve. The distinction matters because a genus-one curve has only a one-dimensional space of regular differentials.

Exclude small orbits

After lifting to characteristic zero, a Galois group acts on the finite collection of covers. A Sylow $p$-subgroup partitions that collection into orbits whose sizes are powers of $p$. If every orbit has size at least $p^h$, then the total number is divisible by $p^h$.

The manuscript therefore rules out a small orbit. Such an orbit would lead, through a descent argument, to a cover defined over a field extension with small $p$-part in its degree. Reduction back to characteristic $p$ then produces curves whose genera and ramification have to satisfy competing constraints.

One key device upgrades an approximate inequality to an exact one. A geometric error is bounded independently of $s$, while the quantity carrying that error is divisible by an increasingly large power of $p$. Once the divisibility step is larger than the possible negative error, a negative value becomes impossible.

The resulting equalities, together with connectedness of a suitable model, force a nontrivial normal $p$-subgroup of $J$. This contradicts $O_p(J)=1$. All orbits must therefore be large, giving the required divisibility. The initial algebraic reduction converts that divisibility back into the blockwise counting equality.

7. Scope, limitations, and open questions

The asserted scope is the full numerical blockwise conjecture. The coefficient field is algebraically closed of characteristic $p$; the paper explains why extending such fields does not change the counts or block assignments. Auxiliary fields used to construct curves are separate from these representation-theoretic coefficient fields.

The result does not classify simple modules, give their dimensions, or provide an efficient way to list them. Nor does equality of two finite cardinalities automatically supply a canonical or equivariant matching. Refined local–global questions retain content beyond this numerical conclusion.

The geometric argument also relies on more than a familiar character formula for one surface. Uniformity as the number of punctures grows, descent with the specified group action, and control of inseparable reduction are essential parts of the claimed advance. Omitting them would obscure what the manuscript actually contributes.

A natural further question is how broadly its divisibility mechanism can be used in other representation-theoretic counting problems. The exposition here reports the stated theorem and its mechanism; it does not resolve those extensions.

8. Sources and further reading

For the original problem, see J. L. Alperin, Weights for finite groups, Proceedings of Symposia in Pure Mathematics 47 (1987), 369–379. Alperin and Fong, Weights for symmetric and general linear groups, Journal of Algebra 131 (1990), 2–22, records early large-family cases.

Knörr and Robinson, Some remarks on a conjecture of Alperin, Journal of the London Mathematical Society 39 (1989), 48–60, supplies the subgroup-chain viewpoint. Späth, A reduction theorem for the blockwise Alperin weight conjecture, Journal of Group Theory 16 (2013), 159–220, represents the simple-group reduction approach.

For the functorial consequence, see Boltje, Bouc and Yılmaz, On Alperin’s conjecture and functorial equivalence of blocks, Mathematische Zeitschrift 312 (2026), Article 112. The source manuscript below contains the precise block-assignment convention and geometric statements.

Source manuscripts

• M1. OpenAI, The Blockwise Alperin Weight Conjecture (September 23, 2026).

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