---
title: Endotrivial Complexes
url: https://www.emergentmind.com/topics/endotrivial-complexes
type: topic
---

# Endotrivial Complexes

Endotrivial complexes are the invertible objects in the bounded homotopy category of \(p\)-permutation modules for a finite group. For a finite group \(G\), a field \(k\) of characteristic \(p\), and the bounded homotopy category
\[
K^b({}_{kG}\mathbf{triv}),
\]
where \({}_{kG}\mathbf{triv}\) denotes the category of finitely generated \(p\)-permutation (equivalently, trivial source) \(kG\)-modules, an endotrivial complex \(C\) is characterized by
\[
C^*\otimes_k C \simeq k[0].
\]
Its equivalence class lies in the Picard group
\[
\mathcal E_k(G)=E_k(G)=\operatorname{Pic}\bigl(K^b({}_{kG}\mathbf{triv})\bigr),
\]
which extends the classical group of endotrivial modules from degree-zero objects to a tensor-triangulated context [2403.04088]. The subject now encompasses a complete classification over fields of prime characteristic, relative variants defined by \(V\)-projectivity, Euler-characteristic and Lefschetz invariants, and an integral generalization over arbitrary commutative Noetherian rings [2403.04088] [2402.08042] [2508.07404] [2605.31128].

## 1. Foundational setup and categorical meaning

The category \(K^b({}_{kG}\mathbf{triv})\) is symmetric monoidal under degreewise tensor product over \(k\), with unit \(k[0]\). If \(C\) is a bounded chain complex of \(p\)-permutation \(kG\)-modules, its dual is
\[
C^*:=\operatorname{Hom}_k(C,k),
\]
and in this category one has
\[
\operatorname{End}_k(C)\cong C^*\otimes_k C.
\]
Accordingly, \(C\) is endotrivial precisely when it is tensor-invertible, or equivalently when
\[
C^*\otimes_k C\simeq k[0].
\]
The group law on \(\mathcal E_k(G)\) is induced by tensor product,
\[
[C_1]+[C_2]=[C_1\otimes_k C_2],
\]
the identity is \([k[0]]\), and the inverse is \([C]^{-1}=[C^*]\) [2309.12138].

Classical endotrivial modules are recovered as complexes concentrated in degree \(0\): if \(M\) is an endotrivial \(kG\)-module, then \(M[0]\) is an endotrivial complex. However, the extension from modules to complexes is genuinely richer. An endotrivial complex need not be quasi-isomorphic to an endotrivial module in degree \(0\); its local structure is encoded by a degree function on \(p\)-subgroups, the \(h\)-mark [2403.04088].

The bounded homotopy category is also the natural home of splendid Rickard theory. If \(C\) is endotrivial, then
\[
\operatorname{Ind}^{G\times G}_{\Delta G}C
\]
is a splendid Rickard autoequivalence of \(kG\), and this construction yields an injective homomorphism
\[
E_k(G)\hookrightarrow S_k(kG),
\]
where \(S_k(kG)\) denotes the group of splendid Rickard autoequivalences modulo homotopy [2309.12138].

## 2. Local invariants and the classification over fields

A central structural theorem is the Brauer-local criterion: a bounded complex \(C\) is endotrivial if and only if, for every \(P\in s_p(G)\), the Brauer construction \(C(P)\) has nonzero homology in exactly one degree, and in the endotrivial case that homology is one-dimensional [2309.12138]. This leads to two local invariants. For an endotrivial complex \(C\), one defines
\[
h_C(P)\in \mathbb Z
\]
as the unique degree with
\[
H_{h_C(P)}(C(P))\neq 0,
\]
and
\[
H_C(P):=H_{h_C(P)}(C(P)),
\]
which is a \(1\)-dimensional \(k[N_G(P)/P]\)-module. These assemble into an injective homomorphism
\[
\Xi:E_k(G)\to \prod_{P\in s_p(G)}\big(\mathbb Z\times \operatorname{Hom}(N_G(P)/P,k^\times)\big).
\]
Projecting to the integer coordinate gives the \(h\)-mark homomorphism
\[
h:E_k(G)\to CF(G,p).
\]

The complete classification theorem states that for every finite group \(G\),
\[
0\to \operatorname{Hom}(G,k^\times)\to E_k(G)\xrightarrow{\,h\,} CF_b(G,p)\to 0
\]
is a split exact sequence of abelian groups. Equivalently,
\[
E_k(G)\cong \operatorname{Hom}(G,k^\times)\oplus CF_b(G,p)
\]
noncanonically. Here \(CF_b(G,p)\) is the group of Borel–Smith superclass functions on the \(p\)-subgroups of \(G\); thus the free part of \(E_k(G)\) is described by Borel–Smith functions, while the torsion is exactly the group of one-dimensional \(kG\)-modules [2403.04088].

For \(p\)-groups the picture simplifies drastically. Since every one-dimensional module over a \(p\)-group in characteristic \(p\) is trivial,
\[
h:E_k(G)\xrightarrow{\sim} CF_b(G).
\]
In fact the paper proves
\[
E_k(-)\cong CF_b(-)
\]
as rational \(p\)-biset functors, so the classification is functorial rather than merely pointwise [2403.04088].

The Borel–Smith conditions are the precise congruence and rank-two relations satisfied by \(h\)-marks. For example, if \(T/S\cong (\mathbb Z/p)^2\), then
\[
h(S)-\sum_{S<Y<T} h(Y)+p\,h(T)=0.
\]
If \(T/S\) is cyclic of order \(p\) for odd \(p\), or cyclic of order \(4\), then
\[
h(S)\equiv h(\widehat S)\pmod 2,
\]
and if \(T/S\) is quaternion of order \(8\), then
\[
h(S)\equiv h(\widehat S)\pmod 4.
\]
These conditions identify exactly which integer-valued superclass functions occur as \(h\)-marks of endotrivial complexes [2403.04088].

A recurrent misconception is that the theory adds nothing beyond endotrivial modules and shifts. The classification shows otherwise: the free part is governed by local integer data on all \(p\)-subgroups, and an endotrivial complex can carry \(h\)-mark information not realizable by any degree-zero module [2403.04088].

## 3. Relative endotrivial complexes

A relative theory replaces projective error terms by \(V\)-projective ones, where a \(kG\)-module or bounded complex \(M\) is \(V\)-projective if
\[
M \mid V\otimes_k N
\]
for some \(N\). For a bounded complex \(C\in Ch^b({}_{kG}\mathbf{triv})\), three notions are introduced:

1. **Weakly \(V\)-endotrivial**:
   \[
   C^* \otimes_k C \simeq k[0]\oplus D,
   \]
   where \(D\) is a bounded chain complex of \(V\)-projective \(kG\)-modules.

2. **Strongly \(V\)-endotrivial**:
   \[
   C^* \otimes_k C \simeq k[0]\oplus D,
   \]
   where \(D\) is a bounded \(V\)-projective \(kG\)-chain complex.

3. **\(V\)-endosplit-trivial**:
   \[
   C^* \otimes_k C \simeq k[0]\oplus M[0],
   \]
   where \(M\) is a \(V\)-projective \(kG\)-module.

These satisfy
\[
V\text{-endosplit-trivial} \implies \text{strongly }V\text{-endotrivial} \implies \text{weakly }V\text{-endotrivial},
\]
and when \(V=0\) all three coincide with ordinary endotriviality [2402.08042].

The relative groups of classes are denoted
\[
wE_k^V(G),\qquad sE_k^V(G),\qquad eE_k^V(G),
\]
with \(eE_k^V(G)\subseteq sE_k^V(G)\). A relatively endotrivial complex has a unique indecomposable relatively endotrivial direct summand, called the cap, and this yields Dade-group-style descriptions for \(sE_k^V(G)\) and \(eE_k^V(G)\) [2402.08042].

The local criterion remains Brauer-theoretic. If \(V\) is a \(p\)-permutation module, absolutely \(p\)-divisible, and
\[
X_V=\{P\in s_p(G)\mid V(P)=0\},
\]
then \(C\) is weakly \(V\)-endotrivial if and only if for all \(P\in X_V\), \(C(P)\) has nonzero homology concentrated in exactly one degree, with the nontrivial homology having \(k\)-dimension one. For \(V\)-endosplit-triviality, the analogous statement runs over all \(P\in s_p(G)\), and for \(P\in X_V\) the local homology must again be one-dimensional [2402.08042].

The \(h\)-mark technology extends as well. For weakly \(V\)-endotrivial complexes,
\[
h_w:wE_k^V(G)\to CF(G,X_V),
\]
and for \(V\)-endosplit-trivial complexes,
\[
h_e:eE_k^V(G)\to CF(G,p).
\]
In both cases the kernel is the module-theoretic torsion subgroup:
\[
\ker(h_w)\cong T_{k,V}(G,S),\qquad \ker(h_e)\cong T_{k,V}(G,S).
\]
Hence \(wE_k^V(G)\) and \(eE_k^V(G)\) are finitely generated abelian, and the relative \(h\)-mark captures the torsion-free part [2402.08042].

The relative theory is not merely a formal variation. The paper shows that the relative setting can be strictly larger than the ordinary one, and it does not fully classify all strongly \(V\)-endotrivial complexes in general [2402.08042].

## 4. Euler characteristic, Lefschetz invariants, and orthogonal units

For an endotrivial complex \(C\), the Lefschetz invariant is
\[
\Lambda(C)=\sum_{i\in\mathbb Z}(-1)^i[C_i]\in T(kG),
\]
where \(T(kG)\) is the trivial source ring. This defines the Lefschetz homomorphism
\[
\Lambda:E_k(G)\to O(T(kG)),
\]
with \(O(T(kG))\) the orthogonal unit group of the trivial source ring [2309.12138].

Under the Boltje–Carman decomposition, the paper identifies \(\Lambda\) explicitly in terms of local invariants:
\[
\Lambda([C])=\bigl(\dim(h_C),\, H(C)\bigr),
\]
where
\[
\dim(f)(H)=(-1)^{f(H)}.
\]
Thus the Euler characteristic of an endotrivial complex consists of the parity pattern of its \(h\)-marks together with its local one-dimensional homology characters [2508.07404].

This connects endotrivial complexes to \(p\)-permutation equivalences and splendid Rickard theory. For a \(p\)-group \(G\), the paper proves that
\[
\Lambda:E_k(G)\to O(T(kG))
\]
is surjective, and deduces that every \(p\)-permutation autoequivalence of a \(p\)-group arises from a splendid Rickard autoequivalence [2403.04088]. More generally, when \(p=2\) and \(k=\mathbb F_2\), \(\Lambda\) is surjective when \(G\) has a Sylow \(2\)-subgroup with fusion controlled by its normalizer, and when \(G\) has dihedral Sylow \(2\)-subgroups. When \(p\) is odd, \(\Lambda\) is surjective if \(G\) has a cyclic Sylow \(p\)-subgroup or is \(p\)-nilpotent, but there are examples of groups of \(p\)-rank \(2\) or greater for which \(\Lambda\) is not surjective [2508.07404].

The kernel is similarly explicit in important cases. For \(p=2\), an endotrivial complex with trivial homology lies in \(\ker(\Lambda)\) if and only if its \(h\)-mark is even-valued. For odd \(p\) and cyclic Sylow \(p\)-subgroups, the criterion is: \(C\in\ker(\Lambda)\) if and only if \(h_C\) is even-valued and, for every nontrivial \(p\)-subgroup \(P\le S\),
\[
h_C(1)\equiv h_C(P)\pmod{2\Phi(S)},
\]
where
\[
\Phi(S)=|\operatorname{Aut}_G(S)|=[N_G(S):C_G(S)].
\]
These results show that the Euler characteristic map is highly structured, but not universally surjective [2508.07404].

## 5. Integral and Noetherian coefficient generalizations

The theory extends from fields to arbitrary commutative Noetherian rings. If \(R\) is a commutative Noetherian ring, let
\[
K(G;R)\coloneqq K_b(\mathsf{perm}(G;R)^\natural),
\]
the bounded homotopy category of finitely generated \(\natural\)-permutation \(RG\)-modules. An endotrivial complex over \(R\) is an invertible object of \(K(G;R)\), so the relevant group is
\[
\operatorname{Pic}(K(G;R)).
\]
This is the Picard group of the derived category of permutation modules for \(G\) over \(R\) [2605.31128].

For connected Noetherian \(R\), the main classification is
\[
0 \to \operatorname{Pic}(R)\times M(G;R)\to \operatorname{Pic}(K(G;R))\xrightarrow{h} \operatorname{CF}^{gl}_b(G;R)\to 0,
\]
a split exact sequence. Equivalently,
\[
\operatorname{Pic}(K(G;R)) = \operatorname{Pic}(R)\times \operatorname{CF}^{gl}_b(G;R)\times M(G;R),
\]
where \(M(G;R)\leq \operatorname{Hom}(G,R^\times)\) is the subgroup of characters \(\varphi\) such that the rank-one module \(R_\varphi\) is \(\natural\)-permutation, and \(\operatorname{CF}^{gl}_b(G;R)\) is the group of fiberwise Borel–Smith functions glued across connected components of the arithmetic fibers [2605.31128].

In the integral case the classification becomes especially clean:
\[
\operatorname{Pic}(K(G;\mathbb Z)) \cong \operatorname{CF}_b(G,\Phi).
\]
Here the line-bundle term \(\operatorname{Pic}(\mathbb Z)\) and the rank-one \(\natural\)-permutation term \(M(G;\mathbb Z)\) vanish, so integral endotrivial complexes are classified purely by Borel–Smith functions on prime-power subgroups [2605.31128].

The global structure is controlled by descent. The paper proves
\[
D(G;R)\xrightarrow{\simeq} \lim_{G/H\in \mathcal O_\Phi(G)^{op}} D(H;R),
\]
and hence
\[
\operatorname{Pic}(K(G;R)) \xrightarrow{\simeq} \lim_{G/H\in \mathcal O_\Phi(G)^{op}} \operatorname{Pic}(K(H;R)),
\]
where \(\mathcal O_\Phi(G)\) is the full subcategory of the orbit category on subgroups of prime-power order. Thus an endotrivial for \(G\) is exactly a restriction-coherent system of endotrivials on prime-power subgroups [2605.31128].

A further structural result is that oriented endotrivial complexes are line bundles on the Balmer spectrum: they become shifts of the tensor unit on an open cover of \(\operatorname{Spc}(K(G;R))\). At the same time, the integral theory is not exhausted by topology: not every endotrivial with integer coefficients arises from a homotopy representation, and for non-nilpotent groups the natural map
\[
\operatorname{Pic}(\mathrm{SH}(G)^c)\to \operatorname{Pic}(K(G;\mathbb Z))
\]
need not be surjective [2605.31128].

## 6. Relations to endotrivial modules, orbit categories, and homotopical models

Endotrivial complexes extend several earlier frameworks for endotrivial modules rather than replacing them. In stable module theory, endotrivial modules are invertible objects in the stable module category, and for cyclic \(p\)-groups and generalized quaternion groups one has
\[
T(G)\cong \operatorname{Pic}(\mathrm{StMod}(kG)).
\]
This identification was computed homotopically by Galois descent and Picard spectral sequences, with explicit results such as
\[
\operatorname{Pic}(\mathrm{StMod}(kQ_{2^n}))\cong C_2\oplus C_4 \qquad (n\ge 4),
\]
showing that endotriviality can already be treated as a problem about Picard groups of stable homotopy theories [2107.06308]. This provides a direct conceptual precursor to the complex-level Picard groups studied later.

Orbit-category and topos-theoretic approaches to endotrivial modules are likewise relevant. For the Sylow-trivial subgroup
\[
T_k(G,P)=\ker\bigl(T_k(G)\to T_k(P)\bigr),
\]
the comparison theorem
\[
\check H^1(U_{G/G},\mathbb G_m)\cong H^1(O_p^\circ(G),k^\times)
\]
identifies Balmer’s Čech-cohomological formula with Grodal’s category-cohomological formula. The same paper also proves
\[
T_k(G,P)\cong \operatorname{Pic}(k),
\]
where \(\operatorname{Pic}(k)\) is the Picard group of the ringed site \((O_p^\circ(G),k)\), so Sylow-trivial modules become invertible \(k\)-linear presheaves on the orbit category [2308.16838]. This is not yet a theory of endotrivial complexes, but it shows that invertibility questions in modular representation theory can be recast as descent and sheaf-theoretic Picard problems.

A complementary geometric model identifies the relative endotrivial module group with prime-to-\(p\) torsion equivariant line bundles on the Brown complex:
\[
T_k(G,P)\xrightarrow{\sim}\operatorname{Tors}_{p'}\operatorname{Pic}^G(\mathcal S_p(G)).
\]
Here \(\mathcal S_p(G)\) is the simplicial complex of non-trivial \(p\)-subgroups. This realization encodes relative endotriviality by equivariant line bundles and weak homomorphisms, and it makes the connection between modular representation theory and equivariant topology explicit [1511.00519].

Two clarifications follow from the later complex theory. First, endotrivial complexes are not merely endotrivial modules placed in degree \(0\); the classification by \(h\)-marks shows that the complex theory has genuinely new free directions [2403.04088]. Second, not every orthogonal unit in the trivial source ring is the Euler characteristic of an endotrivial complex; the failure of surjectivity of \(\Lambda\) in odd characteristic and in groups of \(p\)-rank at least \(2\) shows that the passage from decategorified \(p\)-permutation data to actual invertible complexes is constrained [2508.07404].

Taken together, these results place endotrivial complexes at the intersection of modular representation theory, biset-functor methods, orbit-category cohomology, tensor-triangular geometry, and equivariant topology. Over fields of characteristic \(p\), they are completely classified by one-dimensional characters and Borel–Smith \(h\)-marks; in the relative setting they admit \(V\)-projective refinements with their own local invariants; and over general Noetherian rings they organize into a genuinely arithmetic Picard theory of permutation-module complexes [2403.04088] [2402.08042] [2605.31128].

Source: https://www.emergentmind.com/topics/endotrivial-complexes