January 01, 1970
REPRESENTATION RINGS OF FUSION SYSTEMS
AND BRAUER CHARACTERS
Thomas Lawrence
2cm2cm Abstract. Let \(\mathop{\mathrm{\mathcal{F}}}\) be a saturated fusion system on a \(p\)-group \(S\). We study the ring \(R(\mathop{\mathrm{\mathcal{F}}})\) of \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters by exploiting a new connection to the modular characters of a finite group \(G\) with \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\). We utilise this connection to find the rank of the \(\mathop{\mathrm{\mathcal{F}}}\)-stable character ring over fields with positive characteristic. We use this theory to derive a decomposition of the regular representation for a fixed basis \(B\) of the ring of complex \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters and give a formula for the absolute value of the determinant of the \(\mathop{\mathrm{\mathcal{F}}}\)-character table with respect to \(B\) (the matrix of the values taken by elements of \(B\) on each \(\mathop{\mathrm{\mathcal{F}}}\)-conjugacy class) for a wide class of saturated fusion systems, including all non-exotic fusion systems, and prove this value squared is a power of \(p\) for all saturated fusion systems.
-1cm-1cm
Introduction
A fusion system \(\mathop{\mathrm{\mathcal{F}}}\) on a \(p\)-group \(S\) is a category with objects \(\text{Ob}(\mathop{\mathrm{\mathcal{F}}}) = \{P \colon P \leq S\}\) and morphism sets \(\text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P, Q)\) consisting of injective group homomorphisms designed to mimic conjugation (see Definitions 1.1 and 2 for a precise exposition). This definition is too general for our purposes, we only work with saturated fusion systems, which are fusion systems with morphism sets satisfying some technical conditions (see Definition 2). From this, we will omit the adjective “saturated" when appropriate.
The basic example of a fusion system is the inner fusion system \(\mathop{\mathrm{\mathcal{F}}}_S(S)\), which is the fusion system with \(\text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P, Q) = \text{Hom}_{S}(P, Q) := \{c_s \in \text{Inn}(S) \colon c_s(P) \leq Q\}\).
Let \(G\) be a finite group and \(S \leq G\) be a \(p\)-subgroup, then we denote the (not necessarily saturated) fusion system with \(\text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P, Q) = \text{Hom}_G(P,Q)\) as \(\mathop{\mathrm{\mathcal{F}}}_S(G)\). We say that \(\mathop{\mathrm{\mathcal{F}}}\) is realised by \(G\) if \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) (see Definition 7). It is key to note that our definition of realised does not require \(S \in \text{Syl}_p(G)\) and we will explicitly mention when \(S\) must be Sylow.
If \(S \in \text{Syl}_p(G)\) then \(\mathop{\mathrm{\mathcal{F}}}_S(G)\) is guaranteed to be saturated (Theorem I.2.3 in [1]) and for every saturated fusion system there is some finite group \(G\) with \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) (see Theorem 3 in [2]).
If \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) and \(S \in \text{Syl}_p(G)\) we say that \(\mathop{\mathrm{\mathcal{F}}}\) is non-exotic, this is because there exist exotic fusion systems \(\mathop{\mathrm{\mathcal{F}}}\) that are saturated but there are no finite groups \(G\) with \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) and \(S \in \text{Syl}_p(G)\) simultaneously.
Two subgroups \(P, Q \leq S\) are \(\mathop{\mathrm{\mathcal{F}}}\)-conjugate if \(\text{Iso}_{\mathop{\mathrm{\mathcal{F}}}}(P, Q) \neq \varnothing\). Similarly, two elements \(x, y \in S\) are \(\mathop{\mathrm{\mathcal{F}}}\)-conjugate if there is an \(\mathop{\mathrm{\mathcal{F}}}\)-isomorphism \(\phi\) with \(\phi(x) = y\). We write the \(\mathop{\mathrm{\mathcal{F}}}\)-conjugacy class of \(x\) as \(x^{\mathop{\mathrm{\mathcal{F}}}}\) and the set of all \(\mathop{\mathrm{\mathcal{F}}}\)-conjugacy classes of elements as \(\text{cl}(\mathop{\mathrm{\mathcal{F}}})\).
A function \(f\) with domain \(S\) is \(\mathop{\mathrm{\mathcal{F}}}\)-stable if \(f|_Q \circ \phi = f|_P\) for all \(\phi \in \text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P, Q)\) and \(P, Q \leq S\). This definition may be applied to both characters and Brauer characters of \(S\).
There has been recent interest (see completion?, Theorem?, unique?, factorisation?, symmetric?, groups?, [3], dimension?, functors?) in exploring the ring \(R(\mathop{\mathrm{\mathcal{F}}})\) of virtual complex \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters of \(S\), which is the Grothendieck completion of the semiring \(R^+(\mathop{\mathrm{\mathcal{F}}})\) of complex \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters. We will expand this study to the ring \(R(\mathop{\mathrm{\mathcal{F}}}, \ell)\) of \(\mathop{\mathrm{\mathcal{F}}}\)-stable \(\ell\)-Brauer characters at a prime \(\ell\).
An element \(\chi \in R^+(\mathop{\mathrm{\mathcal{F}}})\) is \(\mathop{\mathrm{\mathcal{F}}}\)-indecomposable if \(\chi \neq \psi+\psi'\) for all \(\psi, \psi' \in R^+(\mathop{\mathrm{\mathcal{F}}})\) (see Definition 10) and we write \(\text{Ind}(\mathop{\mathrm{\mathcal{F}}})\) for the set of \(\mathop{\mathrm{\mathcal{F}}}\)-indecomposable characters. Unlike ordinary character theory, \(R(\mathop{\mathrm{\mathcal{F}}})\) is not always freely generated by Ind\((\mathop{\mathrm{\mathcal{F}}})\) (see Example A.2 in unique?, factorisation?). Despite this, we still know that \(\text{rk}_{\mathop{\mathrm{\mathbb{Z}}}}(R(\mathop{\mathrm{\mathcal{F}}})) =|\text{cl}(\mathop{\mathrm{\mathcal{F}}})|\) (see Corollary 2.2 of completion?, Theorem?) therefore, there is a basis \(B\) of \(R(\mathop{\mathrm{\mathcal{F}}})\) of size \(|\text{cl}(\mathop{\mathrm{\mathcal{F}}})|\).
The basis \(B\) is defines the \(\mathop{\mathrm{\mathcal{F}}}\)-character table with respect to \(B\): \(X_{B}(\mathop{\mathrm{\mathcal{F}}}) := (\chi(x_K))_{\chi \in B, K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})}\) where the \(x_K \in K\) is some \(\mathop{\mathrm{\mathcal{F}}}\)-conjugacy class representative. We refer the curious reader to unique?, factorisation? and symmetric?, groups? for some in-depth calculations of \(X_B(\mathop{\mathrm{\mathcal{F}}})\) for various examples.
Based on unpublished calculations, Jason Semeraro has conjectured the following:
Conjecture 1. (Semeraro) Let \(\mathop{\mathrm{\mathcal{F}}}\) be a fusion system on a \(p\)-group \(S\) and \(B\) be any basis of \(R(\mathop{\mathrm{\mathcal{F}}})\), then \(|\text{det}(X_B(\mathop{\mathrm{\mathcal{F}}}))|^2 = \displaystyle\prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} |C_S(x_K)|\), where \(x_K \in K\) is a fully \(\mathop{\mathrm{\mathcal{F}}}\)-centralised conjugacy class representative.
We note that this conjecture implies that this determinant is independent from our choice \(B\) up to sign, a fact we prove in Lemma 41. Additionally, because \(x_K\) is chosen to be fully \(\mathop{\mathrm{\mathcal{F}}}\)-centralised, \(|C_S(x_K)| \geq |C_S(y)|\) for all \(y \in x_K^{\mathop{\mathrm{\mathcal{F}}}}\) so this product is independent of our choice of \(x_K\).
To provide some motivation for this conjecture we will quickly prove it for the inner fusion system \(\mathop{\mathrm{\mathcal{F}}}_S(S)\): Characters are class functions on \(S\), so \(R(S) = R(\mathop{\mathrm{\mathcal{F}}}_S(S))\) and \(X_{\text{Ind}(\mathop{\mathrm{\mathcal{F}}}_S(S))}(\mathop{\mathrm{\mathcal{F}}}_S(S))\) is just the character table for \(S\). So, we have that \[|\text{det}(X_{\mathop{\mathrm{\mathcal{F}}}_S(S)}(\mathop{\mathrm{\mathcal{F}}}_S(S)))|^2 = \text{det}\left(X_{\mathop{\mathrm{\mathcal{F}}}_S(S)}\overline{X_{\mathop{\mathrm{\mathcal{F}}}_S(S)}^T}\right) = \prod_{K \in \text{cl}(S)} |C_S(x_K)|\] by column orthogonality. Combined with the fact that this determinant does not depend on our choice of \(B\) (Lemma 41), Conjecture A holds for \(\mathop{\mathrm{\mathcal{F}}}_S(S)\).
In this paper we show that this conjecture holds for all non-exotic fusion systems and transitive fusion systems:
Theorem 1. (Proposition 43, Theorem 39, Corollary 52, Theorem 20) Conjecture A holds when \(\mathop{\mathrm{\mathcal{F}}}\) is transitive or \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) with \(S \in \text{Syl}_p(G)\) for some finite group \(G\). Additionally, if Conjecture A holds for \(\mathop{\mathrm{\mathcal{F}}}_1, \mathop{\mathrm{\mathcal{F}}}_2\), then it holds for \(\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\).
We were unable to solve this conjecture for all fusion systems, but have managed to prove a weakened version:
Theorem 2. (Corollary 52) Given any fusion system \(\mathop{\mathrm{\mathcal{F}}}\) on a \(p\)-group \(S\), \(|\text{det}(X_B(\mathop{\mathrm{\mathcal{F}}}))|^2\) is a power of \(p\).
The proof of this theorem is arguably more interesting than the theorem itself. We introduce a new symmetry between \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters for \(\mathop{\mathrm{\mathcal{F}}}\) realised by a group \(G\) and Brauer characters (characters of modular representations) of \(G\). For some prime \(p\) we will write the \(p\)-Brauer characters of \(G\) to refer to the characters of \(p\)-modular representations of \(G\), \(p\) will be omitted if the specific prime is unimportant.
In some heuristic sense the \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters behave as if they were the \(p'\)-Brauer characters of some group \(G\). This connection is made clear when \(|G| = p^aq^b\) and \(\mathop{\mathrm{\mathcal{F}}}\) is the fusion system realised by \(G\) over its Sylow \(p\)-subgroup \(S\): the \(\mathop{\mathrm{\mathcal{F}}}\)-indecomposable characters are the irreducible \(q\)-Brauer characters restricted to \(S\) (see Proposition 22).
We may frame these results as a generalisation of the theory of \(\pi\)-partial characters (see Chapter 3 in [4]) in the case that \(\pi\) is every prime dividing \(|G|\) except \(p\). The theory in [4] only holds for \(\pi\)-seperable groups, but our methods will work for any group and any fusion system.
There are analogues of the decomposition and Cartan matrices (Definition 23, Definition 30) and the associated projective indecomposables (Definition 24). We refer the reader to Chapter 2 of [5] for the necessary background on these objects in the context of modular representation theory.
Using this machinery we describe the ring of \(\mathop{\mathrm{\mathcal{F}}}\)-stable Brauer characters:
Theorem 3. (Proposition 46, Theorem 50) Let \(\mathop{\mathrm{\mathcal{F}}}\) be any fusion system on a \(p\)-group \(S\), \(\ell\) a prime. Let \(R(\mathop{\mathrm{\mathcal{F}}}, \ell)\) be the ring of \(\mathop{\mathrm{\mathcal{F}}}\)-stable \(\ell\)-Brauer characters, then \(R(\mathop{\mathrm{\mathcal{F}}}, \ell) \cong R(\mathop{\mathrm{\mathcal{F}}})\) as rings if \(p \neq \ell\) and \(R(\mathop{\mathrm{\mathcal{F}}}, \ell) \cong \mathop{\mathrm{\mathbb{Z}}}\) if \(p = \ell\).
We use this theorem to prove Theorem A by utilising the isomorphism between representation rings to show that the rows of \(X_B(\mathop{\mathrm{\mathcal{F}}})\) are linearly independent “mod \(\ell\)" for all primes \(\ell \neq p\). (see Definition 47 for how this is made rigorous). Combined with the fact that \(\text{det}(X_B(\mathop{\mathrm{\mathcal{F}}}))^2 \in \mathop{\mathrm{\mathbb{Z}}}\) (see Corollary 38), we have that \(\text{det}(X_B(\mathop{\mathrm{\mathcal{F}}}))\) is a power of \(p\). The refinement of this result to the full statement of Conjecture A for non-exotic fusion systems is where the parallels with modular character theory are fully utilised. A corollary to this approach (Corollary 51) is that any basis of \(R(\mathop{\mathrm{\mathcal{F}}})\) explicitly gives a basis of \(R(\mathop{\mathrm{\mathcal{F}}}, \ell)\) for \(\ell \neq p\) by reducing \(B\)”mod \(\ell\)".
Additionally, we are able to elaborate on a recent conjecture (Conjecture 2.18 in symmetric?, groups?) asking if all \(\mathop{\mathrm{\mathcal{F}}}\)-indecomposable characters appear as subcharacters of the regular character \(\rho_S\) of \(S\). This was shown to be false in [3] but we prove that the decomposition of \(\rho_S\) with respect to a basis \(B\) is still well behaved:
Proposition 1. (Proposition 29) Let \(\rho_S\) denote the regular character of \(S\). Let \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) for some \(G\) and \(B\) a basis for \(R(\mathop{\mathrm{\mathcal{F}}})\). For \(\chi \in \text{Irr}(G)\) we write the decomposition of \(\chi|_S\) over \(B\) as \(\sum_{\psi \in B} d^{B,G}_{\chi\psi}\psi\) for some \(d^{B,G}_{\chi\psi} \in \mathop{\mathrm{\mathbb{Z}}}\). Let \(\Phi^{B,G}_{\psi} := \sum_{\chi \in \text{Irr}(G)}d^{B,G}_{\chi\psi}\chi \in R(G)\), then we have \[\rho_S = \sum_{\psi \in B} \frac{\Phi^{B,G}_{\psi}(1)}{[G:S]}\psi\] And that these coefficients are integers.
I - An overview of \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters
Definition 1. A fusion system \(\mathop{\mathrm{\mathcal{F}}}\) over a \(p\)-group \(S\) is a category with \(\text{Ob}(\mathop{\mathrm{\mathcal{F}}}) = \{P \colon P \leq S\}\) and morphism sets \(\text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P, Q)\) consisting of injective group homomorphisms such that \[\text{Hom}_S(P,Q) := \{\phi \in \text{Hom}(P, Q) \colon \phi = c_s|_P \text{ for some } s \in S\} \subseteq \text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P,Q)\] Furthermore, each \(\phi \in \text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P, Q)\) decomposes as an isomorphism followed by an inclusion.
If \(\text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P,Q) = \text{Hom}_{G}(P,Q)\) for all \(P,Q \leq S\) and \(S \in \text{Syl}_p(G)\), then we say that \(\mathop{\mathrm{\mathcal{F}}}\) is realised by \(G\) and write \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\). If \(\text{Iso}_{\mathop{\mathrm{\mathcal{F}}}}(P, Q) \neq \varnothing\) we say that \(P\) and \(Q\) are \(\mathop{\mathrm{\mathcal{F}}}\)-conjugate.
Definition 2. A fusion system \(\mathop{\mathrm{\mathcal{F}}}\) on a \(p\)-group \(S\) is saturated if every subgroup \(P \leq S\) is \(\mathop{\mathrm{\mathcal{F}}}\)-conjugate to some \(Q \leq S\) with \(\text{Aut}_S(Q) := \text{Hom}_S(Q,Q) \in \text{Syl}_p(\text{Aut}_{\mathop{\mathrm{\mathcal{F}}}}(Q))\) and such that for each \(Q' \leq S\), \(\phi \in \text{Iso}_{\mathop{\mathrm{\mathcal{F}}}}(Q', Q)\) then there exists a map \(\overline{\phi} \in \text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(N_{\phi}, S)\) that restricts to \(\phi\), where \(N_{\phi} := \{n \in N_S(Q) \colon \phi \circ c_n \circ \phi^{-1} \in \text{Aut}_S(Q)\}\).
A reminder that all fusion systems we work with are assumed to be saturated and we will omit mentioning that they are.
Definition 3. Given an \(s \in S\), the \(\mathop{\mathrm{\mathcal{F}}}\)-conjugacy class of \(s\) is the set \[s^{\mathop{\mathrm{\mathcal{F}}}} := \{s' \in S \colon \exists P, Q \text{ such that there exists a } \phi \in \text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P, Q) \text{ with } \phi(s') = s\}\] We will write \(\text{cl}(\mathop{\mathrm{\mathcal{F}}})\) for the set of \(\mathop{\mathrm{\mathcal{F}}}\)-conjugacy classes. We will adopt the notation \(x_K\) to stand for an arbitrary representative of \(K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})\).
Definition 4. For a finite group \(G\), we define \(R^+(G)\) to be the semiring of characters of \(G\), and \(R(G)\) to be the ring of (virtual) characters, which is the Grothendieck completion of \(R^+(G)\).
We will also write \(\text{cf}(G)\) for the set of complex valued class functions on \(G\) and \(\text{cf}(\mathop{\mathrm{\mathcal{F}}})\) for the set of \(\mathop{\mathrm{\mathcal{F}}}\)-stable class functions on \(S\), defined as follows:
Definition 5. Let \(f \in \text{cf}(S)\). Then \(f\) is \(\mathop{\mathrm{\mathcal{F}}}\)-stable if for each \(P \leq S\) and \(\phi \in \text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P, S)\), \(f|_P = f|_Q \circ \phi\).
Lemma 6. A class function \(f\) is \(\mathop{\mathrm{\mathcal{F}}}\)-stable if and only if \(f(x) = f(y)\) for all \(x \in y^{\mathop{\mathrm{\mathcal{F}}}}\).
Proof. See Lemma 1.3 in unique?, factorisation?. ◻
Since characters of \(S\) are class functions, this definition of \(\mathop{\mathrm{\mathcal{F}}}\)-stability may be applied. It also follows that the product and sum of two \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters is again \(\mathop{\mathrm{\mathcal{F}}}\)-stable. We will denote the semiring of \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters as \(R^+(\mathop{\mathrm{\mathcal{F}}})\). We call the Grothendieck completion of \(R^+(\mathop{\mathrm{\mathcal{F}}})\) the ring of (virtual) \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters and denote it as \(R(\mathop{\mathrm{\mathcal{F}}})\).
Definition 7. Let \(\mathop{\mathrm{\mathcal{F}}}\) be a fusion system on \(S\). We say \(\mathop{\mathrm{\mathcal{F}}}\) is realised by \(G\) if \(S \leq G\) and for all \(P, Q \leq S\) \[\text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P, Q) = \{\phi \colon P \rightarrow Q \colon \exists x \in G \text{ such that } \phi = c_x|_P\}\] We will write \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) to indicate this.
Lemma 8. For any fusion system \(\mathop{\mathrm{\mathcal{F}}}\) on \(S\), there exists some finite group \(G\) such that \(\mathop{\mathrm{\mathcal{F}}}\) is realised by \(G\).
Proof. See Theorem 3 in [2]. ◻
If there exists a \(G\) with \(S \in \text{Syl}_p(G)\) such that \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\), then \(\mathop{\mathrm{\mathcal{F}}}\) is non-exotic.
Lemma 9. If \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) for some \(G\), then for any \(\chi \in R(G)\), \(\chi|_S \in R(\mathop{\mathrm{\mathcal{F}}})\).
Proof. Clearly \(\chi|_S \in R(S)\), all that remains is to show that \(\chi|_S\) is \(\mathop{\mathrm{\mathcal{F}}}\)-stable. Because \(\mathop{\mathrm{\mathcal{F}}}\) is realised by \(G\), \(\text{Hom}_{\mathop{\mathrm{\mathcal{F}}}}(P,Q) = \text{Hom}_{G}(P,Q)\) so \(s' \in s^{\mathop{\mathrm{\mathcal{F}}}} \iff s' \in s^G\) for all \(s, s' \in S\). Since \(\chi \in \text{cl}(G)\), we have that for any two \(s, s' \in S\) with \(s' \in s^G\) then \(\chi|_S(s) = \chi|_S(s')\). ◻
Definition 10. If \(\chi \in R^+(\mathop{\mathrm{\mathcal{F}}})\) cannot be written as a sum of two other elements in \(R^+(\mathop{\mathrm{\mathcal{F}}})\) we say that \(\chi\) is \(\mathop{\mathrm{\mathcal{F}}}\)-indecomposable. We write the set of \(\mathop{\mathrm{\mathcal{F}}}\)-indecomposable characters as \(\text{Ind}(\mathop{\mathrm{\mathcal{F}}})\).
Proposition 11. We have \(\langle \text{Ind}(\mathop{\mathrm{\mathcal{F}}}) \rangle_{\mathop{\mathrm{\mathbb{Z}}}} = R(\mathop{\mathrm{\mathcal{F}}})\) and \(\langle \text{Ind}(\mathop{\mathrm{\mathcal{F}}}) \rangle_{\mathop{\mathrm{\mathbb{C}}}} = \text{cf}(\mathop{\mathrm{\mathcal{F}}})\).
Proof. Lemma 2.1 in completion?, Theorem?. ◻
Unfortunately, in contrast to ordinary character theory of groups, the representation ring \(R(\mathop{\mathrm{\mathcal{F}}})\) is not freely generated by the \(\mathop{\mathrm{\mathcal{F}}}\)-indecomposable characters (see Example A.2 in unique?, factorisation?). However, we still have some information about the rank of \(R(\mathop{\mathrm{\mathcal{F}}})\):
Theorem 12. Viewed as free abelian group, the rank of \(R(\mathop{\mathrm{\mathcal{F}}})\) is \(|\text{cl}(\mathop{\mathrm{\mathcal{F}}})|\).
Proof. See Corollary 2.2 of completion?, Theorem?. ◻
Corollary 13. \(R(\mathop{\mathrm{\mathcal{F}}})\) is freely generated by \(\text{Ind}(\mathop{\mathrm{\mathcal{F}}})\) if and only if \(|\text{Ind}(\mathop{\mathrm{\mathcal{F}}})| = |\text{cl}(\mathop{\mathrm{\mathcal{F}}})|\).
Proof. See Corollary 2.9 in unique?, factorisation?. ◻
Finally, we will need the following definition:
Definition 14. For \(s \in S\), we say that \(s\) is fully \(\mathop{\mathrm{\mathcal{F}}}\)-centralised when \(|C_S(s)| \geq |C_S(s')|\) for all \(s' \in s^{\mathop{\mathrm{\mathcal{F}}}}\).
II - Product fusion systems and \(\mathop{\mathrm{\mathcal{F}}}\)-character tables
Given a fusion system \(\mathop{\mathrm{\mathcal{F}}}\) on \(S\), and a \(\mathop{\mathrm{\mathbb{Z}}}\)-basis \(B\) of \(R(\mathop{\mathrm{\mathcal{F}}})\), we define the \(\mathop{\mathrm{\mathcal{F}}}\)-character table with respect to \(B\) as the matrix \((X_B(\mathop{\mathrm{\mathcal{F}}}))_{\psi \in B, K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} = \psi(x_K)\). Note that this matrix does not depend on our choice of \(x_K \in K\) because \(\psi\) is \(\mathop{\mathrm{\mathcal{F}}}\)-stable.
We show that if \(\mathop{\mathrm{\mathcal{F}}}\) is a minimal counterexample to Conjecture A, \(\mathop{\mathrm{\mathcal{F}}}\) cannot be a product \(\mathop{\mathrm{\mathcal{F}}}=\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\) of two strictly smaller fusion systems.
Definition 15. Let \(\mathop{\mathrm{\mathcal{F}}}_i\) be a fusion system over \(S_i\) for \(i = 1,2\), then the product fusion system \(\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\) is a fusion system over \(S_1 \times S_2\) with \[\text{Hom}_{\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2}(P,Q) = \{(\phi_1, \phi_2)|_P \colon \phi_i \in \text{Hom}_{\mathop{\mathrm{\mathcal{F}}}_i}(P_i, Q_i), (\phi_1, \phi_2)(P) \leq Q\}\] Such that \(P, Q \leq S_1 \times S_2\) and \(P_i, Q_i\) denoting the projections of \(P, Q\) to \(S_i\).
This definition is from Theorem I.6.6 in [1], which also proves that \(\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\) is saturated over \(S_1 \times S_2\).
Lemma 16. If \((p_1, p_2) \in (q_1, q_2)^{\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2} \iff p_1 \in {q_1}^{\mathop{\mathrm{\mathcal{F}}}_1}\) and \(p_2 \in {q_2}^{\mathop{\mathrm{\mathcal{F}}}_2}\).
Proof. By definition, if there is an \(\mathop{\mathrm{\mathcal{F}}}_1\times\mathop{\mathrm{\mathcal{F}}}_2\)-isomorphism sending \((p_1, p_2) \mapsto (q_1, q_2)\) it is of the form \((\phi_1, \phi_2) \in \text{Iso}_{\mathop{\mathrm{\mathcal{F}}}_1}(P_1, Q_1)\times\text{Iso}_{\mathop{\mathrm{\mathcal{F}}}_2}(P_2, Q_2)\) where \(P_i \leq S_i\) containing \(p_i\) and likewise for \(Q_i\) and \(q_i\). So \(p_i \in {q_i}^{\mathop{\mathrm{\mathcal{F}}}_1}\) with \(\phi_i\) being a map in \(\mathop{\mathrm{\mathcal{F}}}_1\) such that \(\phi_i(p_i) = q_i\).
If we now assume \(p_i \in {q_i}^{\mathop{\mathrm{\mathcal{F}}}_i}\) with \(\phi_i\) being the isomorphism mapping \(p_i\) to \(q_i\). Define \(P_i, Q_i\) as before, then \((\phi_1, \phi_2) \in \text{Iso}_{\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2}((P_1, P_2), (Q_1, Q_2))\), so \((p_1, p_2) \in (q_1, q_2)^{\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2}\). ◻
Lemma 17. Let \(\mathop{\mathrm{\mathcal{F}}}_i\) be a fusion system on \(S_i\) for \(i = 1,2\), then \((x_1,x_2)\) is fully \(\mathop{\mathrm{\mathcal{F}}}=\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\)-centralised if and only if \(x_i\) is fully \(\mathop{\mathrm{\mathcal{F}}}_i\)-centralised.
Proof. It is clear that \(|C_{S_1 \times S_2}((x_1, x_2))| = |C_{S_1}(x_1) \times C_{S_2}(x_2)|\). So if \(x_i\) is fully \(\mathop{\mathrm{\mathcal{F}}}_i\)-centralised, then \(|C_{S_1 \times S_2}((x_1, x_2))| =|C_{S_1}(x_1)||C_{S_2}(x_2)| \geq |C_{S_1}(x_1')||C_{S_2}(x_2')| = |C_{S_1 \times S_2}((x_1', x_2'))|\) for all \(x_i' \in x_i^{\mathop{\mathrm{\mathcal{F}}}}\). Because \((x_1, x_2)^{\mathop{\mathrm{\mathcal{F}}}} = x_1^{\mathop{\mathrm{\mathcal{F}}}_1} \times x_2^{\mathop{\mathrm{\mathcal{F}}}_2}\) by Lemma 16, we have our result. ◻
With the structure of \(\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\) known, we are able to describe \(R(\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2)\):
Theorem 18. If \(\chi \in \text{R}(\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2)\) then \(\chi = \chi_1\chi_2\) with \(\chi_i \in R(\mathop{\mathrm{\mathcal{F}}}_i)\). So \(R(\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2) \cong R(\mathop{\mathrm{\mathcal{F}}}_1) \times R(\mathop{\mathrm{\mathcal{F}}}_2)\).
Proof. We first assume \(\chi_i\) are \(\mathop{\mathrm{\mathcal{F}}}_i\)-stable. If \((p_1, p_2) \in (q_1, q_2)^{\mathop{\mathrm{\mathcal{F}}}_1\times\mathop{\mathrm{\mathcal{F}}}_2}\) then \(p_1 \in q_1^{\mathop{\mathrm{\mathcal{F}}}_1}\) and \(p_2 \in q_2^{\mathop{\mathrm{\mathcal{F}}}_2}\). Then since \(\chi_1, \chi_2\) are \(\mathop{\mathrm{\mathcal{F}}}_1\)-stable and \(\mathop{\mathrm{\mathcal{F}}}_2\)-stable respectively, we have \(\chi(p_1, p_2) = \chi_1(p_1)\chi_2(p_2) = \chi_1(q_1)\chi_2(q_2) = \chi(q_1, q_2)\). Thus \(\chi\) is \(\mathop{\mathrm{\mathcal{F}}}_1\times\mathop{\mathrm{\mathcal{F}}}_2\)-stable.
Now, take an \(\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\)-stable character \(\chi\). Since \(\chi \in R(S_1 \times S_2)\) we can decompose it as \(\chi = \chi_1\chi_2\) with \(\chi_i \in R(S_i)\).
Let \((p_1, p_2) \in (q_1, q_1)^{\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2}\), so \(\chi(p_1, p_2) = \chi(q_1, q_2)\) and \(\chi_1(p_1)\chi_2(p_2) = \chi_1(q_1)\chi_2(q_2)\). So if we set \(p_2 = 1\) we have \(\chi_1(p_1)\chi_2(1) = \chi_1(q_1)\chi_2(1) \Rightarrow \chi_1(p_1) = \chi_1(q_1)\). Since \((p_1, 1) \in (q_1,1)^{\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2} \iff p_1 \in q_1^{\mathop{\mathrm{\mathcal{F}}}_1}\) we can conclude that \(\chi_1\) is \(\mathop{\mathrm{\mathcal{F}}}_1\)-stable. An identical argument holds for \(\chi_2\) by setting \(p_1 = 1\) and the result is shown. ◻
Corollary 19. If \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\) and \(B_i\) a basis for \(R(\mathop{\mathrm{\mathcal{F}}}_i)\), define \(B_1B_2 := \{\psi\mu \colon \psi \in B_1, \mu \in B_2\}\) then \(B_1B_2\) is a basis for \(R(\mathop{\mathrm{\mathcal{F}}})\) and \(X_{B_1B_2}(\mathop{\mathrm{\mathcal{F}}}) = X_{B_1}(\mathop{\mathrm{\mathcal{F}}}_1) \otimes X(\mathop{\mathrm{\mathcal{F}}}_2)_{B_2}\), where we use \(\otimes\) to denote the Kronecker product of matrices.
Proof. By Theorem 18 we have for any \(\chi \in R(\mathop{\mathrm{\mathcal{F}}})\) that \(\chi = \chi_1\chi_2\) for some \(\chi_i \in R(\mathop{\mathrm{\mathcal{F}}}_i)\), hence \[\chi = \chi_1\chi_2 = \left(\sum_{\psi\in B_1} \alpha_{\psi}\psi\right)\left(\sum_{\mu \in B_2} \alpha_{\mu}\mu\right) = \sum_{\psi \in B_1}\sum_{\mu \in B_2} \alpha'_{\psi\mu}\psi\mu\] So \(B_1B_2\) is a generating set for \(R(\mathop{\mathrm{\mathcal{F}}})\). We have the equality of matrices: \[\begin{align} (X_{B_1}(\mathop{\mathrm{\mathcal{F}}}_1) \otimes X_{B_2}(\mathop{\mathrm{\mathcal{F}}}_2))_{(\psi, \mu) \in B_1B_2, (s_1, s_2) \in S_1 \times S_2} &= \psi(s_1)\mu(s_2) \\ &= \psi\mu(s_1, s_2) \\ &= (X_{B_1B_2}(\mathop{\mathrm{\mathcal{F}}}))_{(\psi,\mu) \in B_1 \times B_2, (s_1, s_2) \in S_1 \times S_2} \end{align}\] And then since the Kronecker product of two full rank matrices is of full rank, we conclude that \(B_1B_2\) is a basis for \(R(\mathop{\mathrm{\mathcal{F}}})\). ◻
Theorem 20. If Conjecture A holds for \(\mathop{\mathrm{\mathcal{F}}}_1, \mathop{\mathrm{\mathcal{F}}}_2\), it holds for \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\).
Proof. We begin by computing \(\text{det}(X_{B_1\otimes B_2}(\mathop{\mathrm{\mathcal{F}}})) = \text{det}(X_{B_1}(\mathop{\mathrm{\mathcal{F}}}_1)) \otimes \text{det}(X_{B_2}(\mathop{\mathrm{\mathcal{F}}}_2))\). The eigenvalues of \(A \otimes B\) is the set of products \(a_ib_j\) where \(a_i\) and \(b_j\) are eigenvalues of \(A, B\) with multiplicity. Hence: \[\text{det}(A \otimes B) = \prod_{i=1}^{\text{dim}(A)}\prod_{j=1}^{\text{dim}(B)}(a_ib_j) = \left(\prod_{i=1}^{\text{dim}(A)} a_i\right)\left(\prod_{i=1}^{\text{dim}(B)} b_i\right) = \text{det}(A)^{\text{dim}(B)}\text{det}(B)^{\text{dim}(A)}\] Thus \[\begin{align} \text{det}(X_{B_1 \otimes B_2}(\mathop{\mathrm{\mathcal{F}}})) = \text{det}(X_{B_1}(\mathop{\mathrm{\mathcal{F}}}_1))^{|\text{cl}(\mathop{\mathrm{\mathcal{F}}}_2)|}\text{det}(X_{B_2}(\mathop{\mathrm{\mathcal{F}}}_2))^{|\text{cl}(\mathop{\mathrm{\mathcal{F}}}_1)|} \end{align}\] Where \(\text{dim}(X_{B_i}(\mathop{\mathrm{\mathcal{F}}})) = |\text{cl}(\mathop{\mathrm{\mathcal{F}}}_i)|\) from Theorem 12. Since we’re assuming Conjecture A holds for \(\mathop{\mathrm{\mathcal{F}}}_i\), we have: \[\begin{align} \text{det}(X_{B_1 \otimes B_2}(\mathop{\mathrm{\mathcal{F}}})) &= \left(\prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}}_1)} |C_{S_1}(x_K)|\right)^{|\text{cl}(\mathop{\mathrm{\mathcal{F}}}_1)|}\left(\prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}}_2)}(|C_{S_2}(x_K)|\right)^{|\text{cl}(\mathop{\mathrm{\mathcal{F}}}_2)|} \\ &= \prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}}_1)}\prod_{J \in \text{cl}(\mathop{\mathrm{\mathcal{F}}}_2)}|C_{S_1}(x_K)||C_{S_2}(x_J)| \\ &= \prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}}_1)}\prod_{J \in \text{cl}(\mathop{\mathrm{\mathcal{F}}}_2)}|C_{S_1 \times S_2}((x_K, x_J))| \end{align}\] Since \(x_K, x_J\) are fully \(\mathop{\mathrm{\mathcal{F}}}_1, \mathop{\mathrm{\mathcal{F}}}_2\)-centralised respectively, \(x_L := (x_K, x_J)\) is fully \(\mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\)-centralised by Lemma 17, combined with \[\prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}}_1)}\prod_{J \in \text{cl}(\mathop{\mathrm{\mathcal{F}}}_2)}|C_{S_1 \times S_2}((x_K, x_J))| = \prod_{L \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})}|C_{S_1 \times S_2}(x_L)|\] the result is shown. ◻
Therefore, if \(\mathop{\mathrm{\mathcal{F}}}\) is a minimal counter-example to the Conjecture A, we cannot have \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_1 \times \mathop{\mathrm{\mathcal{F}}}_2\), as one of \(\mathop{\mathrm{\mathcal{F}}}_1, \mathop{\mathrm{\mathcal{F}}}_2\) would be a smaller counter-example.
III - Analogy with Brauer characters
We fix a fusion system \(\mathop{\mathrm{\mathcal{F}}}\) over \(S\), fix a a basis \(B\) of \(R(\mathop{\mathrm{\mathcal{F}}})\) and a \(G\) that realises \(\mathop{\mathrm{\mathcal{F}}}\). While this will not directly impact any of ourse results, we remark that this dependence on \(G\) can usually be mitigated. For example, if \(\mathop{\mathrm{\mathcal{F}}}\) is constrained then then there are models for \(\mathop{\mathrm{\mathcal{F}}}\) which are unique up to isomorphism (see Definition I.4.8 and Theorem I.4.9 in [1]). More generally we will take \(G\) to have minimal order.
Definition 21. For a finite group \(G\) we write cl\(_p(G)\) and cl\(_{p'}(G)\) for the set of \(p\)-conjugacy classes and \(p'\)-conjugacy classes of \(G\).
The entire view of \(\mathop{\mathrm{\mathcal{F}}}\)-stable character theory being “\(p'\)-Brauer character theory" was initially motivated by the following result:
Proposition 22. Let \(|G| = p^aq^b\), \(S \in \text{Syl}_p(G)\) and \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\). Write \(\text{IBr}_q(G)\) for the set of irreducible \(q\)-Brauer characters of \(G\). Then \(\text{Ind}(\mathop{\mathrm{\mathcal{F}}})\) is the set of Brauer lifts of \(\text{IBr}_q(G)|_S := \{\psi|_S \colon \psi \in \text{IBr}_q(G)\}\).
Proof. Note that \(\text{cl}_{q'}(G) = \text{cl}_{p}(G)\) and \(\text{cl}(\mathop{\mathrm{\mathcal{F}}}) = \text{cl}_{p}(G) \cap S\). Now take \(\psi \in \text{IBr}_q(G)\). Because \(S\) is a \(q'\)-group, we have a bijection \(\psi \mapsto \psi'\) between \(\text{IBr}_q(S)\) and \(\text{Irr}(S)\) given by Brauer lifting (see Theorem 2.12 in [5], Theorem 43.ii in [6]). It is clear that \(\psi|_S' = \psi'|_S\).
Because restrictions of Brauer characters are Brauer characters and \(S\) is a \(q'\)-group we have \(\psi'|_S \in R(S)\) for \(\psi \in \text{IBr}_q(G)\), furthermore \(\psi'\) is invariant on \(\text{cl}_{q'}(G) = \text{cl}_p(G)\) so we have \(\psi'|_S \in R(\mathop{\mathrm{\mathcal{F}}})\).
As all Sylow \(p\)-subgroups are conjugate, \(x \in \bigcup_{K \in \text{cl}_{p}(G)} K\) is conjugate to some \(\tilde{x} \in S\). We define an additive map \(\phi\) extending \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters to class functions on \(\text{cl}_{q'}(G)\) by \(\phi(\chi)(x) := \chi(\tilde{x})\), Since \(\chi\) is \(\mathop{\mathrm{\mathcal{F}}}\)-stable this map is well defined and does not depend on the choice of \(\tilde{x}\) and it is clear that \(\phi\) is a ring homomorphism.
We show that the \(\chi \in \text{Ind}(\mathop{\mathrm{\mathcal{F}}})\) is the restriction to \(S\) of the lift of an irreducible \(q\)-Brauer character of \(G\). Firstly we note that \(\chi\) is the restriction of a non-virtual \(q\)-Brauer character \(\psi\) of \(G\) because \(\psi|_S\) is still non-virtual, and then because Brauer lifting is a ring homomorphism that maps \(\text{IBr}_q(S) \rightarrow \text{Irr}(S)\) bijectively, lifts of non-virtual \(q\)-Brauer characters of \(S\) are non-virtual characters.
If \(\chi \in \text{Ind}(\mathop{\mathrm{\mathcal{F}}})\) with \(\chi = \psi'_1|_S+\psi'_2|_S\) for two \(\psi_1, \psi_2 \in \text{IBr}_q(G)\) then this immediately contradicts the \(\mathop{\mathrm{\mathcal{F}}}\)-indecomposablity of \(\chi\), hence \(\chi = \psi'|_S\) for a single \(\psi \in \text{IBr}_q(G)\). ◻
Definition 23. Let \(\mathop{\mathrm{\mathcal{F}}}\) be a fusion system on \(S\). The decomposition matrix of \(\mathop{\mathrm{\mathcal{F}}}\) with respect to \(B\) and \(G\) is the matrix \[(D_{B,G}(\mathop{\mathrm{\mathcal{F}}}))_{\chi \in \text{Irr}(G), \psi \in B} := \langle \chi|_S, \psi \rangle\] The decomposition numbers with respect to \(B\) and \(G\) are defined as \(d^{B,G}_{\chi\psi} = D_{B,G}(\mathop{\mathrm{\mathcal{F}}})_{\chi\psi}\).
Note that since \(\chi|_S \in R^+(\mathop{\mathrm{\mathcal{F}}})\) by Lemma 9, we know that the decomposition numbers are integers, and will be positive integers when \(B \subseteq \text{Ind}(\mathop{\mathrm{\mathcal{F}}})\).
Definition 24. For \(\psi \in B\), we write \(\displaystyle\Phi^{B,G}_{\psi} := \sum_{\chi \in \text{Irr}(G)} d^{B,G}_{\chi\psi}\chi \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})\). We define the matrix \((P_{B,G})_{\psi \in B, K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} := \Phi^{B,G}_{\psi}(x_K)\).
\(B\) and \(G\) are fixed so we will omit them from our notation and write \(D(\mathop{\mathrm{\mathcal{F}}}) := D_{B,G}(\mathop{\mathrm{\mathcal{F}}})\), \(d_{\chi\psi} := d^{B,G}_{\chi\psi}\), \(\Phi^{B,G}_{\psi} := \Phi_{\psi}\), \(P_{B,G} := P\). These \(\Phi_{\psi}\) are the fusion theoretic versions “projective indecomposable associated to \(\psi\)" from modular character theory. One can quickly see that \(\Phi_{\psi} \in R(G)\) because the decomposition numbers are integers. We list some useful properties of these characters:
Lemma 25. Let \(X(\mathop{\mathrm{\mathcal{F}}})\) be the character table of \(\mathop{\mathrm{\mathcal{F}}}\) and \(\Delta := \text{Diag}_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})}(|C_G(x_K)|)\). Then \(\overline{P^T}X(\mathop{\mathrm{\mathcal{F}}}) = \Delta\), and therefore \(X(\mathop{\mathrm{\mathcal{F}}})\) and \(P\) are both of full rank.
Proof. Let \(g \in G\), \(s \in S\), then: \[\begin{align} \sum_{\psi \in B} \overline{\Phi_{\psi}(g)}\psi(s) &= \sum_{\psi \in B}\sum_{\chi \in \text{Irr}(G)} d_{\chi\psi}\overline{\chi(g)}\psi(s) = \sum_{\chi \in \text{Irr}(G)} \overline{\chi(g)}\sum_{\psi \in B}d_{\chi\psi}\overline{\psi(s)} = \sum_{\chi \in \text{Irr}(G)} \overline{\chi(g)}\chi(s) \\ &= \delta_{g^{G}s^{G}}|C_G(x)| \end{align}\] By column orthogonality. Writing out \(\overline{P^T}Q\) gives us: \[\begin{align} (\overline{P^T}X(\mathop{\mathrm{\mathcal{F}}}))_{K, K' \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} &= \sum_{\psi \in B}\overline{\Phi_{\psi}(x_K)}\psi(x_{K'}) = \delta_{K,K'}|C_G(x_K)| \\ &\Rightarrow \overline{P^T}X(\mathop{\mathrm{\mathcal{F}}}) = \text{Diag}_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})}(|C_G(x_K)|) = \Delta \end{align}\] Since \(|C_G(x_K)| \neq 0\) for any \(K\), \(\Delta\) is of full rank. We know that \(P\) is square by Theorem 12, therefore \[|\text{cl}(\mathop{\mathrm{\mathcal{F}}})| = \text{rk}(\Delta) \leq \text{min}(\text{rk}(P), \text{rk}(X(\mathop{\mathrm{\mathcal{F}}}))) \leq |\text{cl}(\mathop{\mathrm{\mathcal{F}}})| \Rightarrow \text{rk}(P) \text{ and } \text{rk}(X(\mathop{\mathrm{\mathcal{F}}})) = |\text{cl}(\mathop{\mathrm{\mathcal{F}}})|\] and both results are proven. ◻
The degree of the associated projective indecomposables for some group \(G\) are always divisible by the size of a Sylow \(p\)-subgroup and they are zero on \(p\)-elements by Corollary 2.14 and Theorem 2.13 in [5] respectively. We have similar results:
Lemma 26. \(\Phi_{\psi}(g) = 0\) whenever \(g\) is not a \(p\)-element of \(G\).
Proof. Let \(g \in G\) such that \(g\) is not a \(p\)-element. Then we have that \(g^G \cap S = \varnothing\), hence \(\sum_{\psi \in B} \overline{\Phi_{\psi}(g)}\psi(s) = 0\) for all \(s \in S\) by Lemma 25. By the linear independence of \(B\), we conclude that \(\Phi_{\psi}(g) = 0\) for all \(\psi\). ◻
Corollary 27. \(|G|_{p'}\) divides \(\Phi_{\psi}(1)\).
Proof. Let \(q \neq p\) be a prime dividing \(|G|\) and \(Q \in \text{Syl}_q(G)\), then \(\langle \Phi_{\psi}|_Q, 1_Q \rangle_Q = \frac{1}{|Q|}\sum_{q \in Q} \Phi_{\psi}|_Q(q)\) which is just \(\frac{\Phi_{\psi}(1)}{|Q|}\) because \(Q \cap S = 1\) and \(\Phi_{\psi}\) is zero everywhere else in \(Q\). Since \(\Phi_{\psi}\) is a virtual character of \(G\), \(\langle \Phi_{\psi}|_Q, 1_Q\rangle \in \mathop{\mathrm{\mathbb{Z}}}\), and so \(|Q|\) divides \(\Phi_{\psi}(1)\).
Because the above argument holds for every \(q \neq p\) dividing \(|G|\), \(|G|_{p'}\) must divide \(\Phi_{\psi}(1)\). ◻
Corollary 28. Let zcf\(_p(G)\) be the set of complex valued class functions \(f\) of \(G\) with \(f(g) = 0\) for all \(g \in G\) that are not \(p\)-elements. Then \(\{\Phi_{\psi}\}_{\psi \in B}\) is a basis for zcf\(_p(G)\).
Proof. By Lemma 26 we know \(\Phi_{\psi} \in \text{zcf}_p(G)\) for all \(\psi \in B\), and by Lemma 25 \(P\) is full rank hence \(\{\Phi_{\psi}\}_{\psi \in B}\) is linearly independent. Combining this with Theorem 12 we have \(\text{rk}(\text{zcf}_p(G)) = |\text{cl}(\mathop{\mathrm{\mathcal{F}}})| = |B| = \text{rk}(\langle \{\Phi_{\psi}\}_{\psi \in B} \rangle)\) and we’re done. ◻
We are now able to prove Proposition D:
Proposition 29. Assume that \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\). Let \(\rho_S\), \(\rho_G\) be the regular characters of \(S\) and \(G\) respectively. Then \(\rho_S = \displaystyle\sum_{\psi \in B} \frac{\Phi_{\psi}(1)}{[G:S]}\psi\) and these coefficients are integers.
Proof. \[\rho_G|_S = \sum_{\chi \in \text{Irr}(G)} \chi(1)\chi|_S = \sum_{\chi \in \text{Irr}(G)}\left(\chi(1)\sum_{\psi \in B}d_{\chi\psi}\psi\right) = \sum_{\psi \in B}\left(\psi\sum_{\chi \in \text{Irr}(G)}d_{\chi\psi}\chi(1)\right) = \sum_{\psi \in B} \psi\Phi_{\psi}(1)\] Now \(\rho_G|_S = [G : S]\rho_S\) thus \[\rho_S = \sum_{\psi \in B}\frac{\Phi_{\psi}(1)}{[G:S]}\psi\] If these coefficients were not integers, then because \(\rho_S \in R(\mathop{\mathrm{\mathcal{F}}})\) there exists unique integers \(a_{\psi}\) giving a decomposition \(\rho_S = \sum_{\psi \in B} a_{\psi}\psi\). Furthermore, \(\mathop{\mathrm{\mathbb{Q}}}\) is flat as a \(\mathop{\mathrm{\mathbb{Z}}}\)-module, so \(B\) is also a basis of \(\mathop{\mathrm{\mathbb{Q}}}\otimes R(\mathop{\mathrm{\mathcal{F}}})\). But then we’d have the linear relation \(0 = \rho_S-\rho_S = \sum_{\psi \in B} (a_{\psi}-\frac{\Phi_{\psi}(1)}{[G:S]})\psi\) in \(\mathop{\mathrm{\mathbb{Q}}}\otimes R(\mathop{\mathrm{\mathcal{F}}})\), which is a contradiction. So \(\frac{\Phi_{\psi}(1)}{[G:S]} = a_{\psi} \in \mathop{\mathrm{\mathbb{Z}}}\). ◻
While we do not use it, we note that this proof generalises Corollary 27. Now we borrow our last object from modular character theory:
Definition 30. We define the Cartan matrix of \(\mathop{\mathrm{\mathcal{F}}}\) with respect to the basis \(B\) of \(R(\mathop{\mathrm{\mathcal{F}}})\) and group \(G\) realising \(\mathop{\mathrm{\mathcal{F}}}\) to be the matrix \(C_{B, G}(\mathop{\mathrm{\mathcal{F}}}) := D_{B, G}(\mathop{\mathrm{\mathcal{F}}})^TD_{B, G}(\mathop{\mathrm{\mathcal{F}}})\). \(C_{B, G}(\mathop{\mathrm{\mathcal{F}}})\) is indexed by \(\psi, \mu \in B\). We define the Cartan numbers with respect to \(B\) and \(G\) to be \(c^{B,G}_{\psi\mu} = C_{B,G}(\mathop{\mathrm{\mathcal{F}}})_{\psi\mu}\).
Remark 31. Notice that \(C_{B,G}(\mathop{\mathrm{\mathcal{F}}})\) is an integer matrix whenever \(D_{B,G}(\mathop{\mathrm{\mathcal{F}}})\) is, hence the Cartan numbers are also integers.
Again, since \(B\) and \(G\) are fixed we omit them from our notation and set \(C(\mathop{\mathrm{\mathcal{F}}}) := C_{B,G}(\mathop{\mathrm{\mathcal{F}}})\), \(c_{\psi\mu} := c^{B,G}_{\psi\mu}\).
Proposition 32. We have \(\langle \Phi_{\psi}, \Phi_{\mu} \rangle_G = c_{\psi\mu}\) for any \(\psi, \mu \in B\).
Proof. \[\begin{align} \langle \Phi_{\psi}, \Phi_{\mu} \rangle_G &= \sum_{\chi \in \text{Irr}(G)}\sum_{\chi' \in \text{Irr}(G)}d_{\psi\chi}d_{\mu\chi'}\langle \chi, \chi' \rangle = \sum_{\chi \in \text{Irr}(G)} d_{\psi\chi}d_{\mu\chi} \\ &= (D^TD)_{\psi\mu} = c_{\psi\mu} \end{align}\] ◻
The Cartan numbers in modular character theory describe how the associative projective indecomposables of \(G\) decompose when restricted to the \(p'\)-elements of \(G\) (see the remark on page 25 of [5]). We have a similar result:
Corollary 33. For all \(\psi \in B\), \(\displaystyle\Phi_{\psi}|_S = \sum_{\mu \in B} c_{\psi\mu}\mu\).
Proof. By definition of the Cartan numbers, \(c_{\psi\mu} = \displaystyle\sum_{\chi \in \text{Irr}(G)} d_{\chi\psi}d_{\chi\mu}\), thus: \[\Phi_{\psi}|_S = \sum_{\chi \in \text{Irr}(G)} d_{\chi\psi}\chi|_S = \sum_{\chi \in \text{Irr}(G)}\sum_{\mu \in B} d_{\chi\psi}d_{\chi\mu}\mu = \sum_{\mu \in B} c_{\psi\mu}\mu\] ◻
Our next goal is to show that the determinant of the Cartan matrix is coprime to \(p\), and then relate that determinant to the determinant of \(X(\mathop{\mathrm{\mathcal{F}}})\). We note that by Corollary 2.18 in [5], that the Cartan matrix of \(p\)-Brauer characters has determinant equal to a power of \(p\).
Lemma 34. \(C(\mathop{\mathrm{\mathcal{F}}})^{-1}\) is the matrix defined by \(\displaystyle(C(\mathop{\mathrm{\mathcal{F}}})^{-1})_{\psi, \mu \in B} := \sum_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} \frac{\psi(x_K)\overline{\mu(x_K)}}{|C_G(x_K)|}\).
Proof. Write \(C'\) for the matrix \(\displaystyle(C')_{\psi, \mu \in B} := \sum_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} \frac{\psi(x_K)\overline{\mu(x_K)}}{|C_G(x_K)|}\). Then \[\begin{align} (C'C(\mathop{\mathrm{\mathcal{F}}}))_{\mu\psi} &= \sum_{\theta \in B} C'_{\mu\theta}C(\mathop{\mathrm{\mathcal{F}}})_{\theta\psi} \\ &= \sum_{\theta \in B}\sum_{K \in\text{cl}(\mathop{\mathrm{\mathcal{F}}})} \frac{\mu(x_K)\overline{\theta(x_K)}}{|C_G(x_K)|}c_{\theta\psi} \\ &= \sum_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})}\frac{\mu(x_K)\sum_{\theta \in B} c_{\theta\psi}\overline{\theta(x_K)}}{|C_G(x_K)|} \\ &= \sum_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} \frac{\mu(x_K)\overline{\Phi_{\psi}|_S(x_K)}}{|C_G(x_K)|} \quad \text{ (applying Corollary \ref{proj32restriction})} \end{align}\] Recall that we write \(\Delta := \text{Diag}_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})}(|C_G(x_K)|)\), then we know from Lemma 25 that \[X(\mathop{\mathrm{\mathcal{F}}})\overline{P^T} = \Delta \Rightarrow (X(\mathop{\mathrm{\mathcal{F}}})\Delta^{-1}\overline{P^T})_{\mu, \psi} = \sum_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} \frac{\mu(x_K)\overline{\Phi_{\psi}(x_K)}}{|C_G(x_K)|} = \delta_{\mu,\psi}\] Since \(x_K \in S\) we may replace \(\Phi_{\psi}\) in the above expression with \(\Phi_{\psi}|_S\), hence \(C'C = I_{|\text{cl}(\mathop{\mathrm{\mathcal{F}}})|}\) and the result is proven. ◻
Lemma 35. Assume \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) and \(S \in \text{Syl}_p(G)\), then \(|G|_{p'}^2C(\mathop{\mathrm{\mathcal{F}}})^{-1}\) is an integer valued matrix.
Proof. For \(\chi \in R(S)\) we define \(\tilde{\chi}\) given by \[\tilde{\chi}(g) := \begin{cases} |G|_{p'}\chi(g) & g \text{ is a p-element} \\ 0 & \text{ otherwise} \end{cases}\] \(\tilde{\chi} \in R(G)\) by Brauer’s characterisation of characters. Hence for any \(\mu, \psi \in B\), we have \(\langle \tilde{\mu}, \tilde{\psi} \rangle_{G} \in \mathop{\mathrm{\mathbb{Z}}}\). Now since \(\tilde{\mu}, \tilde{\psi}\) are \(0\) on anything that isn’t a \(p\)-element: \[\langle \tilde{\mu}, \tilde{\psi} \rangle_G = \sum_{K \in \text{cl}(G)} \frac{\tilde{\mu}(x_K)\overline{\tilde{\psi}(x_K)}}{|C_G(x_K)|} =\sum_{K \in \text{cl}(G), K \cap S \neq \varnothing}\frac{\tilde{\mu}(x_K)\overline{\tilde{\psi}(x_K)}}{|C_G(x_K)|}\] All Sylow \(p\)-subgroups are conjugate in \(G\), so we are free to choose our \(G\)-conjugacy class representatives \(x_K\) to lie in \(S\): \[\begin{align} \sum_{K \in \text{cl}(G), K \cap S \neq \varnothing}\frac{\tilde{\mu}(x_K)\overline{\tilde{\psi}(x_K)}}{|C_G(x_K)|} &= \sum_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})}\frac{\tilde{\mu}(x_K)\overline{\tilde{\psi}(x_K)}}{|C_G(x_K)|} \\ &= |G|_{p'}^2 \sum_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})}\frac{\mu(x_K)\overline{\psi(x_K)}}{|C_G(x_K)|} \\ &= |G|_{p'}^2(C(\mathop{\mathrm{\mathcal{F}}})^{-1})_{\mu\psi} \in \mathop{\mathrm{\mathbb{Z}}} \end{align}\] And the result is shown. ◻
Lemma 36. If \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) with \(S \in \text{Syl}_p(G)\), then \(\text{det}(C)\) is coprime to \(p\).
Proof. Using the notation from the proof of Lemma 35, define \(M\) with \((M)_{\mu, \psi \in B} = \langle \tilde{\mu}, \tilde{\psi} \rangle_G\), which is \(|G|^2_{p'}C(\mathop{\mathrm{\mathcal{F}}})^{-1}\) by Lemma 35. So \(C(\mathop{\mathrm{\mathcal{F}}})M = |G|^2_{p'}I_{|\text{cl}(\mathop{\mathrm{\mathcal{F}}})|}\). Since \(M\) is in integer matrix, it has an integer determinant. Furthermore, \(C(\mathop{\mathrm{\mathcal{F}}})\) is an integer matrix by Remark 31 and thus also has an integer determinant. Therefore \[\text{det}(C(\mathop{\mathrm{\mathcal{F}}}))= \frac{|G|_{p'}^{2|\text{cl}(\mathop{\mathrm{\mathcal{F}}})|}}{\text{det}(M)} \in \mathop{\mathrm{\mathbb{Z}}}\Longrightarrow \text{det}(C(\mathop{\mathrm{\mathcal{F}}})) \text{ is coprime to } p\] ◻
Lemma 37. For any fusion system \(\mathop{\mathrm{\mathcal{F}}}\) over \(S\), we have that \[\displaystyle|\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))|^2 = \frac{\prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} |C_G(x_K)|}{\text{det}(C(\mathop{\mathrm{\mathcal{F}}}))}\]
Proof. We surpress our notation and write \(X, D, C\) for \(X(\mathop{\mathrm{\mathcal{F}}}), D(\mathop{\mathrm{\mathcal{F}}}), C(\mathop{\mathrm{\mathcal{F}}})\) in this proof.
Let \(\chi \in \text{Irr}(G)\) and \(K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})\), then: \[(DX)_{\chi K} = \sum_{\mu \in B} D_{\chi\mu}X_{\mu K} = \sum_{\mu \in B} d_{\chi \mu}\mu(x_K) = \chi|_S(x_K)\] So by Lemma 25, we have that \((\overline{(DX)^T}DX)_{ij} = \delta_{ij}|C_G(x_{K_j})|\). Therefore \[\text{det}(\overline{(DX)^T}DX) = \prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})}|C_G(x_K)|\] Finally \[\begin{align} \text{det}(\overline{(DX)^T}DX) &= \text{det}(\overline{D^T}D)\text{det}(\overline{X^T}X) = \text{det}(D^TD)|\text{det}(X)|^2 =\text{det}(C)|\text{det}(X)|^2 \\ &\Rightarrow |\text{det}(X)|^2 = \frac{\prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} |C_G(x_K)|}{\text{det}(C)} \end{align}\] ◻
Corollary 38. We have that \(|\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))|^2\) is an integer.
Proof. Since \(X(\mathop{\mathrm{\mathcal{F}}})\) is a matrix of character values, \(X(\mathop{\mathrm{\mathcal{F}}}) \in M_{|\text{cl}(\mathop{\mathrm{\mathcal{F}}})|}(\mathcal{R})\) where \(\mathcal{R} \subset \mathop{\mathrm{\mathbb{C}}}\) is the ring of algebraic integers. So by Lemma 37 and Remark 31, \(|\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))|^2 \in \mathop{\mathrm{\mathbb{Q}}}\cap \mathcal{R} = \mathop{\mathrm{\mathbb{Z}}}\). ◻
IV - Application of the analogy and modular \(\mathop{\mathrm{\mathcal{F}}}\)-stable characters.
In this section we build towards a proof of the following theorem:
Theorem 39. Let \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) with \(S \in \text{Syl}_p(G)\), then \(|\text{det}(X_B(\mathop{\mathrm{\mathcal{F}}}))|^2 = \displaystyle\prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} |C_G(x_K)|_{p}\).
Remark 40. Assuming \(\mathop{\mathrm{\mathcal{F}}}= \mathop{\mathrm{\mathcal{F}}}_S(G)\) with \(S \in \text{Syl}_p(G)\), we have \(|C_G(x_K)|_{p} = |C_S(x_K)|\) with \(x_K\) fully centralised in \(\mathop{\mathrm{\mathcal{F}}}\) by the remark following Definition I.2.4 in [1] and that \(x_K\) is fully \(\mathop{\mathrm{\mathcal{F}}}_S(G)\)-centralised \(\iff C_S(x_K) \in \text{Syl}_p(C_G(x_K))\). So this theorem is equivalent to Conjecture A for non-exotic fusion systems.
Lemma 41. For any two bases \(B,B'\) of \(R(\mathop{\mathrm{\mathcal{F}}})\), \(|\text{det}(X_B(\mathop{\mathrm{\mathcal{F}}}))|^2 = |\text{det}(X_{B'}(\mathop{\mathrm{\mathcal{F}}}))|^2\).
Proof. Given two bases \(B, B'\) of \(R(\mathop{\mathrm{\mathcal{F}}})\), there is a change of basis matrix \(M \in GL(|B|, \mathop{\mathrm{\mathbb{Z}}})\) such that \(MX_B(\mathop{\mathrm{\mathcal{F}}}) = X_{B'}(\mathop{\mathrm{\mathcal{F}}})\), hence \(\text{det}(M)\text{det}(X_B(\mathop{\mathrm{\mathcal{F}}})) = \pm\text{det}(X_{B}(\mathop{\mathrm{\mathcal{F}}})) = \text{det}(X_{B'}(\mathop{\mathrm{\mathcal{F}}}))\) and therefore \(|\text{det}(X_B(\mathop{\mathrm{\mathcal{F}}}))|^2 = |\text{det}(X_{B'}(\mathop{\mathrm{\mathcal{F}}}))|^2\). ◻
We give more motivation to Conjecture A by proving it for transitive fusion systems directly:
Definition 42. We say that a fusion system \(\mathop{\mathrm{\mathcal{F}}}\) on \(S\) is transitive if there are only two \(\mathop{\mathrm{\mathcal{F}}}\)-conjugacy classes of \(S\) (and these two classes must be \(\{1\}\) and \(S-\{1\}\)).
Proposition 43. Let \(\mathop{\mathrm{\mathcal{F}}}\) be a transitive fusion system on \(S\), then Conjecture A holds.
Proof. Set \(B = \text{Ind}(\mathop{\mathrm{\mathcal{F}}})\), by Lemma 41 it is enough to show Conjecture A for this basis.
By Lemma 2.16 in unique?, factorisation?, \(R(\mathop{\mathrm{\mathcal{F}}})\) is freely generated by \(\text{Ind}(\mathop{\mathrm{\mathcal{F}}})\) when \(\mathop{\mathrm{\mathcal{F}}}\) is transitive, so by Corollary 13 we have \(|\text{Ind}(\mathop{\mathrm{\mathcal{F}}})| = |\text{cl}(\mathop{\mathrm{\mathcal{F}}})| = 2\). It is clear to see that \(\text{Ind}(\mathop{\mathrm{\mathcal{F}}}) = \{1_S, \rho_S-1_S\}\) where \(\rho_S\) is the regular character of \(S\) and \(1_S\) is the trivial character, hence \[X := X_{\text{Ind}(\mathop{\mathrm{\mathcal{F}}})}(\mathop{\mathrm{\mathcal{F}}}) = \begin{pmatrix} 1 & 1 \\ |S|-1 & -1 \end{pmatrix}\] So \(|\text{det}(X)|^2 = |S|^2\). Now we label the two \(\mathop{\mathrm{\mathcal{F}}}\)-conjugacy classes \(K_1 = \{1\}, K_2 = S-K_1\). Because \(S\) is a \(p\)-group, \(\mathcal{Z}(S)\) is non trivial. Therefore \(K_2 \cap \mathcal{Z}(S) \neq \varnothing\), so a fully \(\mathop{\mathrm{\mathcal{F}}}\)-centralised representative of \(K_2\) is central. Hence \(\prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} |C_S(x_K)| = |S|^2\) and we are done. ◻
Following Lemma 41, we fix a basis \(B\) and omit it from our notation. To continue further we will need to introduce \(\mathop{\mathrm{\mathcal{F}}}\)-stable modular characters:
Definition 44. For a given prime \(\ell\), we write \(R^+(\mathop{\mathrm{\mathcal{F}}}, \ell)\) to denote the subsemiring of \(R^+(S, \ell)\) consisting of \(\mathop{\mathrm{\mathcal{F}}}\)-stable \(\ell\)-Brauer characters. As before, we write \(R(\mathop{\mathrm{\mathcal{F}}}, \ell)\) for the Grothendieck completion of \(R^+(\mathop{\mathrm{\mathcal{F}}}, \ell)\).
We remark that we do not identify \(R^+(S, \ell)\) with \(R^+(S)\) when \(\ell \neq p\) as the fact that \(R^+(S, \ell)\) consists of functions into a field with positive characteristic will be of interest to us.
Definition 45. Recalling Definition 21, we will write cf\(_{p'}(G)\) for the set of functions \(\text{cl}_{p'}(G) \rightarrow \overline{\mathop{\mathrm{\mathbb{F}}}}_p\).
Proposition 46. If \(\mathop{\mathrm{\mathcal{F}}}\) is a fusion system on a \(p\)-group \(S\), then \(R(\mathop{\mathrm{\mathcal{F}}}, p) = \langle 1_S \rangle\) where \(1_S\) is the trivial \(p\)-Brauer character.
Proof. By Corollary 2.10 in [5], we have that \(|\text{IBr}_p(G)| = |\text{cl}_{p'}(G)|\), combined with the fact that \(S\) is a \(p\)-group, we have that \(|\text{IBr}_p(S)| = |\text{cl}_{p'}(S)| = 1\), and so we have \(\text{IBr}_q(S) = \{1_S\}\). This character is obviously \(\mathop{\mathrm{\mathcal{F}}}\)-stable, and so we have our result. ◻
Definition 47. For a given prime \(\ell\), let \(\mathcal{M}_{\ell}\) be a maximal ideal of the algebraic integers \(\mathcal{R}\) with \(\ell\mathop{\mathrm{\mathbb{Z}}}\subseteq \mathcal{M}_{\ell}\). Denote the canonical surjection \(\mathcal{R} \rightarrow \mathcal{R}/\mathcal{M}_{\ell}\) by \(\pi_{\ell}\). Given a function \(f\) that maps into \(\mathcal{R}\), we abuse notation slightly and write \(\pi_{\ell}(f)\) for the composition \(\pi_{\ell} \circ f\).
From now on we take \(\ell \neq p\) and build up to proving the rest of Theorem 39.
Lemma 48. Let \(G\) be a finite group. For \(K \in \text{cl}_{\ell'}(G)\), let \(i_K \in cf(G)\) be the indicator function for \(K\). Then \(\pi_{\ell}(i_K)|_S \in \langle \text{IBr}_{\ell}(G)|_S \rangle_{\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}}\).
Proof. By Lemma 2.4, Theorem 1.19 in [5] we have that \(\text{IBr}_q(G)\) is linearly independent over \(\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}\). Thus \(\text{rk}\langle \text{IBr}_{\ell}(G) \rangle_{\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}} = |\text{cl}_{\ell'}(G)| = \text{rk}(\text{cf}_{\ell'}(G)) \Rightarrow \text{rk}\langle \text{IBr}_{\ell}(G) \rangle_{\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}} = \text{cl}_{\ell'}(G)\), the result then follows upon restricting to \(S\). ◻
Lemma 49. Let \(\{f_i \colon R_1 \rightarrow R_2\}_{i=1}^n\) be a set of ring homomorphisms with \(R_1\) a commutative ring and \(R_2\) an integral domain. If the \(f_i\) are distinct, they are linearly independent over \(R_2\).
Proof. See Lemma 5.2.2 in [7]. ◻
Theorem 50. Let \(\mathop{\mathrm{\mathcal{F}}}\) be any fusion system over a \(p\)-group \(S\). For \(\ell \neq p\), \(\text{rk}_{\mathop{\mathrm{\mathbb{F}}}_{\ell}}(R(\mathop{\mathrm{\mathcal{F}}}, \ell)) = |\text{cl}(\mathop{\mathrm{\mathcal{F}}})|\).
Proof. For any \(s \in S\) we define the evaluation map \(e_s \colon R(\mathop{\mathrm{\mathcal{F}}}, \ell) \otimes \overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}} \rightarrow \overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}\) with \(e_s(\psi) = \psi(s)\). It is clear that these maps are ring homomorphisms.
Take a set of \(\mathop{\mathrm{\mathcal{F}}}\)-conjugacy class representatives \(x_K\) and consider the set \(\mathscr{E} := \{e_{x_K} \colon K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})\}\). We aim to show that each \(e_{x_K}\) is distinct, so \(\mathscr{E}\) is linearly independent over \(\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}\) by Lemma 49.
By Lemma 8, there exists a finite group \(G\) such that \(\mathop{\mathrm{\mathcal{F}}}\) is realised by \(G\). Consider the set \(\text{IBr}_{\ell}(G)|_S\), these are \(\ell\)-Brauer characters of \(S\) by Lemma 2.2 in [5]. They are also \(\mathop{\mathrm{\mathcal{F}}}\)-stable because \(\mathop{\mathrm{\mathcal{F}}}\) is realised by \(G\) and so are contained in \(R(\mathop{\mathrm{\mathcal{F}}}, \ell)\).
Again, because \(\mathop{\mathrm{\mathcal{F}}}\) is realised by \(G\) and \(\ell \neq p\), we have for all \(G\)-classes \(K\) of elements of \(S\) that \(K \cap S \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})\). Hence by Lemma 48 the indicator functions \(\pi_{\ell}(i_K)\) for \(K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})\) are in the \(\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}\)-span of \(\text{IBr}_{\ell}(G)|_S\), and are therefore in \(R_v(\mathop{\mathrm{\mathcal{F}}}, \ell) \otimes \overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}\).
Because \(e_{x_K}\) are indicator functions, we have \(e_{x_K}(\pi_{\ell}(i_{K'})) = \delta_{K,K'}\) for any \(K, K' \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})\) hence each \(e_{x_K}\) is distinct and \(\mathscr{E}\) is linearly independent over \(\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}\). Since \(R(S, \ell)\) is a finite dimensional \(\mathop{\mathrm{\mathbb{F}}}_{\ell}\)-algebra, \(R(\mathop{\mathrm{\mathcal{F}}}, \ell) \otimes \overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}\) is a finite dimensional \(\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}\)-algebra. Now we have: \[\text{rk}_{\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}}(\langle\mathscr{E}\rangle) = |\text{cl}(\mathop{\mathrm{\mathcal{F}}})| \leq \text{rk}_{\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}}(\text{Hom}(R(\mathop{\mathrm{\mathcal{F}}}, {\ell}) \otimes \overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}, \overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}})) = \text{rk}_{\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}}(R(\mathop{\mathrm{\mathcal{F}}}, {\ell}) \otimes \overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}) = \text{rk}_{\mathop{\mathrm{\mathbb{F}}}_{\ell}}(R(\mathop{\mathrm{\mathcal{F}}}, {\ell}))\] So \(\text{rk}_{\mathop{\mathrm{\mathbb{F}}}_{\ell}}(R(\mathop{\mathrm{\mathcal{F}}}, \ell)) \geq |\text{cl}(\mathop{\mathrm{\mathcal{F}}})|\). Recalling Definition 47, we also have \(R(\mathop{\mathrm{\mathcal{F}}}, \ell) \leq \langle \pi_{\ell}(i_k) \colon K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}}) \rangle\), which has a rank of \(|\text{cl}(\mathop{\mathrm{\mathcal{F}}})|\). Therefore, \(\text{rk}_{\mathop{\mathrm{\mathbb{F}}}_{\ell}}(\mathop{\mathrm{\mathcal{F}}}, \ell) = |\text{cl}(\mathop{\mathrm{\mathcal{F}}})|\) as desired. ◻
Corollary 51. \(\pi_{\ell}\) induces a bijection \(B \mapsto \pi_{\ell}(B)\) and \(\pi_{\ell}(B)\) is linearly independent over \(\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}\).
Proof. Take \(\chi \in \text{IBr}_{\ell}(G)\), we may Brauer lift (see Theorem 43.ii in [6]) \(\chi\) to a \(\chi' \in R(G)\). Then because \(\chi'|_S\) is \(\mathop{\mathrm{\mathcal{F}}}\)-stable we may apply Proposition 11 to write \(\chi'|_S = \sum_{\psi \in B} c_{\psi}\psi\) for some \(c_{\psi} \in \mathop{\mathrm{\mathbb{Z}}}\).
Because each \(c_{\psi} \in \mathop{\mathrm{\mathbb{Z}}}\subset \mathcal{R}\), \(\pi_{\ell}(c_{\psi})\) is well defined. By Theorem 43 in [6] we know that \(\pi_{\ell}(\chi') = \chi\). So \(\pi_{\ell}(\chi'|_S) = \chi|_S = \sum_{\psi \in B} \pi_{\ell}(c_{\psi})\pi_{\ell}(\psi)\).
By Lemma 48 we have that \(\langle \text{IBr}_{\ell}(G)|_S \rangle_{\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}} = R(\mathop{\mathrm{\mathcal{F}}}, {\ell}) \otimes \overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}\), in particular: \[R(\mathop{\mathrm{\mathcal{F}}}, {\ell}) \otimes \overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}} = \langle \text{IBr}_{\ell}(G)|_S \rangle_{\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}} = \langle \pi_{\ell}(B) \rangle_{\overline{\mathop{\mathrm{\mathbb{F}}}_{\ell}}}\] Thus \(\pi_{\ell}(B)\) spans \(R(\mathop{\mathrm{\mathcal{F}}}, {\ell}) \otimes \overline{\mathop{\mathrm{\mathbb{F}}}}_{\ell}\) as a \(\overline{\mathop{\mathrm{\mathbb{F}}}}_{\ell}\)-vector space, which has rank \(|\text{cl}(\mathop{\mathrm{\mathcal{F}}})| = |B|\), so \(|B| = |\pi_{\ell}(B)|\) and \(\pi_{\ell}(B)\) is linearly independent. ◻
Proof of Theorem 39:
By Corollary 38 we know that \(|\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))|^2\) is an integer. Viewing an element of \(B\) as a tuple in \(\mathop{\mathrm{\mathbb{C}}}^{|\text{cl}(\mathop{\mathrm{\mathcal{F}}})|}\), then the rows of \(X(\mathop{\mathrm{\mathcal{F}}})\) are
elements of \(B\), so Corollary 51 implies that \(\pi_{\ell}(\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))) \neq
0\) for all \(\ell \neq p\). Thus \(|\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))|^2 \in \bigcap_{\ell \neq p} \mathop{\mathrm{\mathbb{Z}}}-\ell\mathop{\mathrm{\mathbb{Z}}}\Rightarrow
|\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))|^2\) is a power of \(p\). So by Lemma 37 and Lemma 36 we have that \(|\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))|^2 = \prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} |C_G(x_K)|_{p}\) and \(\text{det}(C) = \prod_{K \in \text{cl}(\mathop{\mathrm{\mathcal{F}}})} |C_G(x_K)|_{p'}\). 0◻
We would be able to use the above proof combined with Lemma 8 to prove Theorem 39 for all fusion systems if it weren’t for the following obstructions:
If \(S\) is not Sylow in \(G\) then we do not have \(C_S(x_K) \in \text{Syl}_p(C_G(x_K))\) for \(x_K\) fully \(\mathop{\mathrm{\mathcal{F}}}\)-centralised, so Remark 40 does not hold.
Lemma 36 requires \(S\) to be Sylow in \(G\) due to a dependence on Lemma 35, where we need all Sylow subgroups to be conjugate in order to choose \(G\)-conjugacy class representatives that lie in \(S\).
Despite this, we do still have Theorem B as a corollary to Theorem 39:
Corollary 52. Let \(\mathop{\mathrm{\mathcal{F}}}\) be any fusion system on \(S\), then \(|\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))|^2\) is a power of \(p\).
Proof. This follows as in Theorem 39: \(|\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))|^2 \in \mathop{\mathrm{\mathbb{Z}}}\) and \(\pi_{\ell}(\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))) \neq 0\) for all primes \(\ell \neq p\) by Corollary 38 and Corollary 51 respectively. Then we again have that \(|\text{det}(X(\mathop{\mathrm{\mathcal{F}}}))|^2 = p^{\alpha}\) for some \(\alpha \in \mathbb{N}\) as before. ◻
If the two obstructions extending this result to the full statement of Conjecture A cannot be removed, then behaviour of the exponent \(\alpha\) as \(\mathop{\mathrm{\mathcal{F}}}\) varies will remain an open problem.
References
Department of Mathematics, Loughborough University, LE11 3TU, United Kingdom
Email: t.lawrence@lboro.ac.uk
Date: 2026-06-16
Keywords: Fusion Systems, Modular Character Theory
2010 Mathematics Subject Classification: 20C20, 20D20, 20C99
Acknowledgements: We thank Jason Semeraro for posing the original problem and numerous helpful discussions and Benjamin Sambale for his helpful comments.↩︎