In this paper
AbstractIntroductionFermionic groupsThe two periodicitiesTorsion in the Picard group of -local -modulesSmith periodicityFrom fermionic groups to -local -modulesProofs of the main theoremsReferencesWorksMathematics
A tale of two periodicities
Abstract
We provide a common explanation for two 4-periodic phenomena arising in the study of connective real K-theory: first, the family of twists of ko over , and second, the torsion in the Picard group of -local -modules. These phenomena have been previously studied, as they lead to 4-periodic patterns in Smith homomorphisms, resp. in the ko-homology of . In this paper, we exhibit both of these 4-periodicities as consequences of an isomorphism of two extraspecial groups of order 32.
Introduction
In topology, many occurrences of the number 8 have the same origin: Bott periodicity in -theory. Quite different-looking 8s, to do with quadratic forms, bordism groups, or Clifford algebras, can be identified by theorems in algebra, geometry, or topology [82].
In prior work joint with Devalapurkar, Liu, Stehouwer, and Thorngren [DDK+25, DDK+26, DKS26], we learned of two different 4-periodic phenomena related to connective -theory. As a parallel to the connections between all these 8s, we wondered whether these two 4-periodicities can be identified.
These two periodicities are:
the 4-periodic family of twists of spin bordism and -theory over , as discussed in [67, 88 HKT20, COSY20, 93 DDK+25, DDK+26], and
the 4-periodic pattern in the -homology of arising from the Picard group of -local stably invertible -modules, as studied by [Yu95, Bru14].
The former periodicity comes from the Thom isomorphism: both -theory and spin bordism have Thom isomorphisms for spin vector bundles, but either theory evaluated on the Thom space of a non-spin vector bundle can be identified with a twisted bordism or -group over the base space. Over , there are four inequivalent twists, corresponding to the trivial rank-zero bundle, , , and , where is the universal bundle. Once we take four copies of , we again have an untwisted Thom isomorphism: admits a spin structure. In quantum physics, this collection of four twisted structures corresponds to a set of four types of fermionic systems: those with a certain symmetry operator that is either unitary or antiunitary and which squares either to or to ; see e.g. the references above for details.
Meanwhile, one can discover the latter periodicity when computing -homology via the Adams spectral sequence. The input data to this computation is a module over the subalgebra of the mod 2 Steenrod algebra. Yu [Yu95] showed that, after localizing at the collection of -Margolis homology isomorphisms (see Definition 2.1), the category of -modules has exactly four isomorphism classes of stably invertible objects up to suspension.
A priori, these two 4-periodic phenomena do not have much in common. We began to wonder whether they were related in prior work of two of us and Stehouwer in [DKS26], where both phenomena played a role in computing twisted -theory of and a corresponding summand of twisted spin bordism.
In this paper, we identify the common origin of both periodicities. Specifically, we define a commutative monoid whose elements are certain extraspecial 2-groups and whose monoid operation, the fermionic tensor product, is a generalization of the central product (see Proposition 1.10). maps to our two periodicities in the following ways.
Taking twisted cohomology defines a homomorphism from to the Picard group of -local -modules.
Passing to the diagonal subgroup (see Example 1.16) defines a twist of over .
We prove in Theorem 0.6 that the torsion subgroup of the group completion of admits surjective maps to the corresponding subgroups in the -twists of spin bordism and the stable -local Picard group, respectively.
Now we go into more detail about our main results. In Example 1.16, we define extraspecial 2-groups of order for all . Heuristically, is defined by taking copies of and copies of , then asking them to commute in the graded sense. Then (Definition 1.21) we define . The fermionic groups within correspond, after taking a twisted group algebra, to the Clifford algebras . There is an isomorphism (see Proposition 1.20), corresponding to the isomorphism . This yields an important relation in , which leads to our first main result.
The map is the twisted cohomology map mentioned above. Theorem 2.10 includes that is an isomorphism, so on its own is also a group completion; we compose with to make the 4-periodicity easier to see.
The torsion summand of constructed in Yu’s theorem [Yu95] (Theorem 2.10) is somewhat mysterious, so it is nice that Theorem 0.1 gives a group-theoretical explanation for it.
In Corollary 1.28, we show that the map sending to the twisting data of the diagonal subgroup defines a monoid homomorphism
Since is an abelian group, by Theorem 0.1 there is a unique group homomorphism
satisfying . This suggests an obvious way to compare these two periodicities: if is the inclusion of the torsion subgroup, we could take the comparison map
However, we show in Proposition 4.3 that this map is not an isomorphism! Instead, it sends .
Nonetheless, these two 4-periodicities are still related, as the fermionic group maps to generators of both s. In § 5 we comment on why the comparison map does not work: there is an “off-by-one error” in that is a suspension of the generator of the torsion subgroup of , and this is the obstruction to being an isomorphism.
Once we correct for this shift, the suspense is over, and we obtain a correspondence between these two periodicities. Specifically, in Lemma 2.12, we construct a projection homomorphism from onto its torsion subgroup by suspending or desuspending a -locally invertible -module until its image in the Picard group is torsion. This suffices to fix the “off-by-one” issue noted above.
Theorem 0.6. Both maps and in the diagram
send to a generator, hence are surjective.
Thus, after applying to desuspend , restricting from to its diagonal subgroup —i.e. applying the map in (5)—identifies the two s we study in this note, hence also the two 4-periodicities they define. As briefly discussed in [DKS26, Remark 5.46], our primary goal in this note is to understand whether these two four-periodic phenomena were related, and Theorem 0.6 offers a strong answer: in this sense, they’re the same 4-periodicity.
It would be interesting to upgrade the correspondence in Theorem 0.6 into a canonically defined group isomorphism —one can use Theorem 0.6 to produce an isomorphism like this, but only after making choices.
Remark 0.8. The four periodicities in this paper yield physical interpretations. Just as Bott periodicity informs the classification of topological insulators and superconductors [65], the four periodicities in this paper give rise to a periodic structure in the classification of more general, interacting versions of these fermionic quantum systems. This structure was originally observed by Queiroz–Khalaf–Stern [81] and two of us, joint with Stehouwer, studied it mathematically in [DKS26].
Outline. We begin in § 2 with a brief introduction to fermionic groups. Though there are already good introductions to fermionic groups in [12, 86, 84 DKS26], in this paper we need a symmetric monoidal -category of discrete fermionic groups, so take a more homotopical viewpoint than those references. We then introduce our fundamental examples (Example 1.16), use them to build the commutative monoid (Definition 1.21) and the diagonal subgroups ((12)), and show that taking the twisting data of the diagonal defines a monoid homomorphism (Corollary 1.28).
In §2.1, we introduce -local stable module categories of -modules, then give the first of our two periodicities, which arises as the subgroup of the -local stable Picard group of -modules. The main result in this subsection is Yu’s calculation [Yu95, Theorem 2.1] (here Theorem 2.10) that this Picard group is isomorphic to .
In §2.2, we compute (Lemma 2.23) and interpret it as giving a 4-periodic sequence of twists of real -theory, hence a -grading on the twisted -cohomology ring of .
In §3, we construct the monoid homomorphism that links these two periodicities together.
Finally, in §4, we prove our main theorems relating these two 4-periodicities: Theorems 0.1 and 0.6 and Proposition 4.3.
Acknowledgments. This paper was born as a spinoff of [DKS26], joint between two of us and Luuk Stehouwer. We would especially like to thank Luuk Stehouwer for many helpful discussions regarding the groups and their role in these two periodicities. We also want to thank Thorger Geiß, Theo Johnson-Freyd, and Noam Nissan for helpful conversations on other aspects of this paper.
CK was supported by the National Science Foundation under Award No. 2602343, the Simons Foundation, and the Killam Trusts during the writing of this paper. CK is based at the Perimeter Institute for Theoretical Physics; research at Perimeter is supported by the Government of Canada through Industry Canada and by the Province of Ontario through the Ministry of Research & Innovation. Any opinions, findings, and conclusions or recommendations expressed in the materials in this work are those of the authors and do not necessarily reflect the views of the National Science Foundation. NPT was supported by the UC President’s Postdoctoral Fellowship.
Usage of AI. One of us consulted Claude during literature review, but we did not use AI to write or prove the results in this paper.
Fermionic groups
Let be an -algebra and let denote the slice -category of spaces with a map . Maps in this -category are continuous maps of spaces that commute with the maps to .
Proposition 1.1 (Ando–Blumberg–Gepner [3 Proposition 6.12]). For an -algebra in spaces, there is a natural symmetric monoidal -category structure on whose value on two objects and is the Cartesian product and external sum
where is the product on .
See Beardsley–Luecke–Morava [17 Remark 5.10] for a little more detail on the relationship between Ando–Blumberg–Gepner’s construction and . Naturality in Proposition 1.1 is with respect to composition with maps of -spaces.
Definition 1.3. Give and the -structures arising from the direct sum of vector bundles, resp. of spin vector bundles. Then, let be the cofiber, in -spaces, of the forgetful map .
Lemma 1.4 ([47 Proposition 1.37 and Lemma 1.41]). As spaces, there is a preferred homotopy equivalence , but this does not extend to an equivalence of -spaces.
Definition 1.5. The category of (discrete)1 fermionic groups is the full sub--category of on the objects such that is connected and -truncated.
The property of being connected and -truncated is preserved by Cartesian products, so the symmetric monoidal structure on restricts to a symmetric monoidal -category structure on .
Remark 1.6. The name fermionic group comes from physics, where these objects may be used to describe the symmetries of quantum systems of fermions; e.g. electrons.
Now that we have a well-behaved category of fermionic groups, let’s see how to actually work with them.
Definition 1.7. Let be a fermionic group. We define the following invariants associated to :
The bosonic group . Since is connected and -truncated, .
The composition
induced by the forgetful map , can be reinterpreted as a class in , which we will also call .
The map classifying the universal bundle induces a map . Proceeding as in the previous bullet point, we obtain a class . This is equivalent data to an isomorphism class of extension of by , which we call .
The pair is called the twisting data of .
Lemma 1.9. The data determine a fermionic group up to isomorphism in the homotopy category of .
Proof. As discussed above, determines up to homotopy, since . The homotopy equivalence mentioned above sends the map to the twisting data , hence determines up to homotopy equivalence. To finish, observe that in the homotopy category of for a space , an isomorphism is a homotopy equivalence of spaces that commutes, up to homotopy, with the maps down to .
We will generally refer to a fermionic group either as the data , or just as when the rest of the data is clear from context. The monoidal product of and in , using the symmetric monoidal structure described above, will be denoted .
Proposition 1.10. Let and be fermionic groups with twisting data and , respectively. Then may be characterized, up to isomorphism in , in either of the following ways.
(1) The fermionic group whose bosonic group is and whose twisting data is
(2) The fermionic tensor product of and as described by Benson [12 §7]: as a set, the direct product , with multiplication
Proof. Part (2) follows from (1) and [DKS26, Lemma 1.12], so we focus on part (1). The monoidal product on is the Cartesian product , which is homotopy equivalent to , so we may take . To compute the twisting data of , it suffices to understand how the -structure on behaves under the (non-) homotopy equivalence , which sends a map to a pair of degree-1 and degree-2 -cohomology classes. In [47 (1.42)] (see also [FH21, (10.9)–(10.11)]), this identification is shown to produce the formula (8). □
Remark 1.13. For a fixed finite group , one can consider the set of isomorphism classes of fermionic groups with bosonic group isomorphic to . By Lemma 1.9, we may equivalently consider the collection of different twisting data over . As a set, this is ; using the grouplike -structure on , this becomes an abelian group, a direct summand of the Brauer group of [DK70, BLM26a]. Explicitly, the product in this group structure on the set has the formula (8).
Warning: we will write elements of as elements of . However, the group operation (8) is not the product of the operations on and .
Example 1.14 (). Recall that is the group of units in the Clifford algebra on one generator , with . Thus and . In both cases, quotienting by gives , with mapping to the generator. We thus obtain two fermionic groups with bosonic group .
Write with .
We choose for both and . This represents the fact that the quotient maps are both surjective.
The class is the extension class of . The extension is split for and nonsplit for , so for and for .
The twisting data for is thus and for is .
Example 1.15. The Conway fermionic group has and , with trivial. Here is the sporadic simple group discovered by Conway in [34, 35], where it is called .1, and is its double cover, called .0 by Conway (ibid.), which is the automorphism group of the Leech lattice. See Johnson-Freyd–Treumann [JFT20, §6] and Albert–Okada–Tachikawa [9] for applications of this fermionic group to questions in physics.
Example 1.16 ([DKS26, Definition 1.36]). Define to be the fermionic tensor product
Thus , and
See [DKS26, (1.40)] for the twisting data of . Define the diagonal subgroup to be the pullback
is a fermionic group with bosonic group and twisting data pulled back from that of .
The first examples are trivial: the inclusions and are isomorphisms. The inclusion is more interesting: one can show that there is an isomorphism sending a rotation to and a reflection to . (Here we are using the description of the fermionic tensor product from part (2) of Proposition 1.10.) is the preimages of and under , so . Therefore is the order-4 group . In other words, is “trivial”: its bosonic group is and its twisting data is .
Proposition 1.20. For , there is an isomorphism of fermionic groups ; restricted to , defines an isomorphism .
Except for the bit about the diagonal subgroups, Proposition 1.20 is proven in [DKS26, Corollary 2.31], which constructs an explicit isomorphism for the case , then takes the fermionic tensor product with to define . The reader can check directly that this choice of restricts to an isomorphism of the diagonal subgroups.
Definition 1.21. Let M be the commutative monoid formed from the set
equipped with the fermionic tensor product .
Lemma 1.23. The map defined by induces an isomorphism
Proof. Proposition 1.20 shows that factors through , so we have a surjective monoid homomorphism as in the lemma statement, and we are done if we can show is injective. Let send . To see that this is a well-defined monoid homomorphism, we check that isomorphic fermionic groups have equivalent images. Isomorphic groups must have identical cardinality, so we have .
For an additional invariant, we can use the following. Given a fermionic group , the fermionic group algebra [12 §7] is defined to be the quotient
where is the unique nonzero element of the kernel of . The homomorphism (see part (2) of Definition 1.7) gives the structure of a -graded algebra over , and the Morita class as a -graded algebra is an invariant of the isomorphism type of .
Albuquerque–Majid [8] show that is isomorphic as a -graded algebra to the Clifford algebra .2 It is standard (see, e.g., [5]) that is -graded Morita equivalent to if and only if mod 8. Thus mod 8 is also an isomorphism invariant of .
We want to show is injective, so we may as well show is injective. Suppose that are such that and mod 8. Then there is some such that . Add this equation to to conclude that or , and thus . That is, and can be identified by applying the relation times. □
Proposition 1.26. For all , the twisting data of , as an element of the abelian group , is the sum of the twisting data of and the twisting data of .
Proof. Let be the twisting data of two fermionic groups whose bosonic groups are both isomorphic to . The addition in the set of homotopy classes of maps to an -space is defined as the pullback of the external sum (see Proposition 1.1) by the diagonal map.
Unspooling this, the sum of and (again, we are using the non-product group law (8)) is the pullback of the twisting data of across the diagonal map .
Now let , , and . By definition, is the pullback
By construction, this map factors through :
Apply the general discussion above to the leftmost square in (17) to see that, as we wanted to show, the twisting data of is the sum of the twisting data of and . □
Corollary 1.28. The map sending to the twisting data of is a homomorphism of commutative monoids .
Corollary 1.29. If , there is an isomorphism of fermionic groups .
Proof. By Proposition 1.20, the homomorphism from Corollary 1.28 identifies the images of and . By Example 1.16, this homomorphism sends , hence the image of only depends on . Combining these two facts, this homomorphism factors through the homomorphism sending .
Example 1.30. and are their own diagonal subgroups: by construction, is the pullback of the inclusion of the diagonal , and the pullback of an isomorphism is an isomorphism. Likewise for . Therefore and are just the twisting data of these two fermionic groups, which we calculated in Example 1.14: and (in the notation of Remark 1.13).
Since the isomorphism type of only depends on by Corollary 1.29, we can write down the four classes.
(0) When , , with trivial twisting data, as we saw in Example 1.16.
(1) When , , as we saw in Example 1.16.
(2) When , one can show that with , e.g. by computing it inside .
(3) When , as in Example 1.16.
The two periodicities
In this section, we introduce the two 4-periodicities that will be the main focus of this paper, as well as some background notions and definitions for them. These manifest as groups isomorphic to , specifically the torsion subgroup of the -local stable Picard group of -modules (Theorem 2.10) and (a nice subgroup of) the twists of -theory over (Lemma 2.23), which we introduce in §2.1 and §2.2 respectively.
Torsion in the Picard group of -local -modules
Let denote the mod 2 Steenrod algebra, the -graded -algebra of natural transformations commuting with the suspension isomorphism. (The grading on is such that a natural transformation increasing the degree of a cohomology class by has degree in .) is generated by the Steenrod squares in degree for each , subject to relations called Adem relations. For any space or spectrum , is naturally a -graded -module, essentially by definition of .
The Cartan formula is the comultiplication map in a commutative Hopf algebra structure on : see [75 Corollary 1 and Appendix 1]. Thus the category of left -modules acquires a symmetric monoidal structure with respect to tensor product over .
For , let denote the subalgebra of generated by with , which is a Hopf subalgebra. We will mostly focus on in this note. Whenever we refer to - or -modules, we implicitly mean in the graded sense: that a degree- element of or acts on a homogeneous element in a way that raises the degree of by .
Definition 2.1 ([7, 72]). Let and inductively define . Thus for . The Adem relations imply that , so for any -module and , one may regard as a chain complex with differential .
In this situation, the -Margolis homology of , denoted , is the homology of the above chain complex.
Since acts in a manner compatible with the grading, the -grading of induces a -grading on . For any , -Margolis homology commutes with direct sums of -modules by definition, and satisfies a Künneth formula with respect to tensor products.
Proposition 2.2 (Margolis [72 Chapter 19, Proposition 17(b)]). For -modules and , and , there is a natural isomorphism
Moreover, [7 §4].
Definition 2.4 (Adams–Margolis [7 Definition 4.1], Margolis [72 Chapter 23]). Let , , and be a map of bounded-below -modules.
The map is a stable isomorphism if its kernel and cokernel are free -modules.
The map is a -local stable isomorphism if is an isomorphism on -Margolis homology.
By [72 Chapter 14, Proposition 1], is a stable isomorphism if and only if there are free -modules and and an -module isomorphism
such that .
Proposition 2.6 (Wall [90 Theorem 2]). The following are equivalent for a map of bounded-below -modules.
is a stable isomorphism.
is both a -local and a -local stable isomorphism.
For , Adams–Margolis [7 Theorem 4.2] gave a similar characterization of stable isomorphisms, but using Margolis homology for a more general class of operations than . Nevertheless, since for any and , every stable isomorphism is a -local stable isomorphism.
For any , it is immediate that the -local equivalences contain all -module isomorphisms and satisfy a two-out-of-three property. Following [70 Proposition 4.1.7.4], to obtain a good theory of localizations, we also need that -local equivalences are stable with respect to tensor products, which follows from the Künneth theorem (Proposition 2.2). Thus we conclude:
Theorem 2.7 (Ricka [83 Proposition 2.6]). There is a symmetric monoidal stable -category which is the localization of the category of bounded-below -modules at the -local equivalences, and the localization functor is symmetric monoidal.
It is also possible to construct the stable module category , the localization of the category of -modules at the stable isomorphisms, as a symmetric monoidal -category in a similar way. See Ricka [83 Proposition 1.9] and Yang [94 §3.5]. We will not need this so will not say much more about it.
Definition 2.8. Let be a symmetric monoidal category.
(1) An object is invertible if there is another object and an isomorphism .
(2) The Picard group of , denoted , is the abelian group of isomorphism classes of invertible objects, with unit the isomorphism class of and the group operation given by .
If is a symmetric monoidal -category, we define to be the Picard group of the homotopy category of .
See [60, 74, 89] for more background on Picard groups, and some example computations.
Example 2.9. The Picard groups of are known for all .
(1) Adams–Priddy [10 Theorem 3.7] constructed an isomorphism ; the unique nonzero torsion class is that of the Joker .
(2) Bhattacharya–Ricka [20 Theorem 4.6] showed that .
(3) Pan–Yan [80 Corollary 4.8] showed that for all , .
(4) Gheorghe–Isaksen–Ricka [GIR18, Theorem 5.6] and Gao–Li [GL24, Theorem 4.3] prove that the Picard groups of the -, resp. -motivic analogues of are both isomorphic to .
We will be interested in the -local Picard group of .
Theorem 2.10 (Yu [Yu95, Theorem 2.1]). There is an isomorphism
such that:
is the -localization of , and
is (which is -local).
The isomorphism induces an isomorphism from the torsion summand of to ; the module in Figure 1 is invertible in and .
Figure 1. By Theorem 2.10, the torsion subgroup of is isomorphic to . In this figure, the -module is in the class . Each module is the unique -module with this property, up to isomorphism, which is -local and has no free -module summands. Though we have but a finite amount of space, each continues upward in the same 4-periodic pattern.
See also Bruner [22 Theorem 7.1 and Proposition 8.2], who gave another proof of Theorem 2.10 and also showed . As far as we know, these are the only values of and for which the isomorphism type of is known. In Figure 1, we draw -local -modules , , , and representing the four torsion classes in .
The first of our two s is the torsion subgroup of . It can be interpreted as a periodicity in the following way: if is a bounded-below, -local -module, then the isomorphism type of , modulo free -module summands, is 4-periodic in . This fact was used to make computations of the -homology of in [Yu95, Theorem A], and of certain twisted -homology groups over in [DKS26, §5].
A corollary of Theorem 2.10 is that there is a preferred projection from onto its torsion subgroup.
Lemma 2.12. The function
defined by suspending or desuspending a -locally invertible -module until it is stably isomorphic to one of the modules is a group homomorphism.
Proof. After applying the isomorphisms
from Theorem 2.10, is identified with the function , which is a homomorphism. □
Smith periodicity
In this subsection, we discuss our second , the group (Lemma 2.23). We discuss how this applies to Picard-group-graded twisted -theory of as well as a relation to families of Smith homomorphisms in bordism theory.
Notation 2.15. Recall the -space from Definition 1.3, and for an -ring spectrum ,3 let be the -group of units of , as defined in May [73 §III.2]. May–Quinn–Ray [73 Lemma IV.2.6] construct an -map , where is the Thom spectrum whose homotopy groups are the spin bordism groups. Using the Atiyah–Bott–Shapiro maps [5], which Joachim [63] showed admit -structures, we then get -maps .
Proposition 2.16. admits a section, and there is an equivalence of -spaces
where is the space with -structure induced from the tensor product of vector bundles.
Madsen–Snaith–Tornehave [77 Lemma 1.13] prove , and the proof of Proposition 2.16 is completely analogous—in particular, Proposition 2.16 is not a new result. See also Antieau–Gepner–Gómez [6 §1], who prove a version of Proposition 2.16 but do not discuss -structures.
We care about because it grades twisted -cohomology, as follows.
Definition 2.18 (Ando–Blumberg–Gepner [3 Definition 1.2]). Let be an -ring spectrum and be a map. The -twisted -cohomology groups of are
where denotes the -module Thom spectrum of , defined in [1 Definition 2.20].
Donovan–Karoubi [DK70] defined twisted -groups associated to a space , , and . See also [64].
Proposition 2.20 (Hebestreit–Sagave [HJ20]). Donovan–Karoubi’s twisted -theory groups are naturally isomorphic to the twisted -cohomology groups, in the Ando–Blumberg–Gepner sense, of the map
Since is an -ring spectrum, the -cohomology groups of a space assemble into a -graded ring, and in fact a -graded ring thanks to Bott periodicity. For a fixed twist , the twisted -groups of do not in general admit a graded ring structure, only a -module structure. This is similar to the structure present (or absent) in cohomology with local coefficients.
On the other hand, if one considers all twists at once, it is possible to multiply: given maps and classes , there is a product , such that [DK70]
has the data of a -graded ring. The subring in gradings is the untwisted -cohomology of with its usual ring structure. This is a cousin of the -grading on ordinary cohomology with local coefficients: see [24, 69, 92, 23 DYY25] for computations using this graded ring structure.
By Proposition 2.16, the -grading splits into two pieces: a grading by and one by . The latter is in general quite mysterious,6 so we will focus on the former.
Lemma 2.23. .
Thus the twisted -cohomology groups of form a -graded ring. It would be interesting to compute this graded ring. Lemma 2.23 was stated without proof in [47 §1.2.3] and [71 Prop. C.37 (1)].
Proof. By Lemma 1.4, there is a homotopy equivalence , so there is a bijection that is not necessarily a group isomorphism
Thus is a group with four elements, so to finish the proof, it suffices to exhibit an element with . For this, we use the fact that, since the map is and the composition
sends (see [47 (1.42)], and the related [DK70, Lemma 7]), we can add classes in by passing through and using the Whitney sum formula: given and in , their sum as elements of is
Thus if is the unique nonzero class, the sum of the twists and is , which is nonzero. □
Remark 2.27. One can show that . Thus we can consider a non-nullhomotopic map and then map to . Such a map is an example of a “higher twist” of , and it would be interesting to better understand how it interacts with the we discuss in this section, either from the point of view of twisted -theory or from that of twisted spin structures. Brooks [21] makes some related computations of higher twists of complex -theory.
One can compute in a few ways. Since is 2-local, we may 2-localize ; then, Adams–Priddy [10 Corollary 1.4] identify with , where denotes the Postnikov 1-connected cover. So we want to understand ; here are two ways forward.
The cofiber sequence induces a long exact sequence in the cohomology of . Bruner–Greenlees [13 §A.9] calculate , and one can attack with the Atiyah–Hirzebruch spectral sequence.
Alternatively, since is a -module, there is a natural isomorphism
Having rephrased this question internal to the category of -module spectra, we can finish it with Baker–Lazarev’s -module Adams spectral sequence [15].7 See [11, 47 Kur25, DKS26, DYY26] for some examples of computations with this spectral sequence over ; in particular, using standard techniques in -modules one can show the spectral sequence computing collapses to imply the claim.
By contrast, Gómez [Góm10] showed that if is any compact Lie group, there are no higher twists of complex -theory over .
The above story using maps to to define a graded ring structure on twisted -cohomology is closely related to the extended gradings on (equivariant) parametrized cohomology discussed in work of Costenoble, Hudson, Tilson, and Waner [CW92, CW16, Cos17, CHT22, CH23, Cos23, CHT24, CH24, CH25a, CH25b, CH25c, CH26] and Beaudry–Lewis–May–Petersen–Tatum [16 BLM+26b], as well as to the permutation twisted cohomology of a finite group as introduced by Balmer–Gallauer [14 §10] and also studied by Gómez and Miller [76, 55, 56].
We finish this subsection with a more geometric description of the isomorphism in Lemma 2.23. Recall that a vector bundle admits a spin structure if its first and second Stiefel–Whitney classes vanish. Twists of spin structure quantify exactly how a vector bundle may fail to be spin.
The following definition, in the special case , is due to B.L. Wang [91 Definition 8.2]. It is also closely related to Kreck’s definition of a normal 1-smoothing [66 §2].
Definition 2.29. Let be a space, , and . An -twisted spin structure on a vector bundle is the data of a map and trivializations of and .
By a trivialization of a cohomology class, we mean a null-homotopy of the representing map to an Eilenberg–Mac Lane space. Also, for fixed, the -twisted spin structures on a vector bundle define a homotopy type; we consider two -twisted spin structures to be equivalent if they belong to the same path component.
Using Lemma 1.4, we will identify with a homotopy class of maps .
Lemma 2.30. Let be a vector bundle, , and . Let be a closed manifold whose tangent bundle has an -twisted spin structure induced by a map . If is a closed submanifold whose mod 2 homology class is Poincaré dual to the mod 2 Euler class of , then has a canonical twisted spin structure for the sum
See also [DYY25, Lemma A.2]. Here is the quotient map. Note that we subtracted off the rank of since classifies rank-zero virtual vector bundles.
Proof. Since the fundamental class of is Poincaré dual to , there is an isomorphism . Then use the Whitney sum formula. □
Lemma 2.30 organizes different kinds of twisted spin structure over the same space into families related by iteratively passing from to .
Example 2.32. Lemma 2.23 implies that for twists of spin structure over , we obtain a 4-periodic family. Specifically, if denotes the tautological bundle, we obtain the following four structures. Under the isomorphism from Lemma 2.23, entry in the list below maps to .
(0) We begin with the trivial twist, where and . A -twisted spin structure on is the data of a spin structure on and a principal -bundle . This is sometimes called a structure.
(1) The next entry in the family is a -twisted spin structure, where . By Peterson [79 §7], this is equivalent to a structure.
(2) Applying (29), our next twist is . By Campbell [25 §7.9], this kind of twisted spin structure is a -structure for .
(3) The next twist is -twisted spin structure, which by Stolz [85 §8] is a structure.
If one applies (29) once more, one gets back to the trivial twist, illustrating that this family is 4-periodic. See [DDK+26, Example 7.8] and references therein. In particular, this family admits a physical interpretation in terms of dimensional reductions of quantum field theories; see e.g. [67, 88 HKT20, COSY20, 93 DDK+25, DDK+26].
Remark 2.33. The periodicity of twists induced by a cyclic subgroup of lifts to a periodicity of iterated Smith homomorphisms, which are roughly speaking the effects on bordism classes of the construction described in Lemma 2.30. This is the context in which this fourfold periodicity for is studied in the references cited above. See [DDK+26, §6] for more information.
Although we discuss them here in terms of the dual to the cohomology Euler class, these homomorphisms are more naturally understood as representing the Euler class in cobordism; see e.g. [DDK+26, Appendix B].
From fermionic groups to -local -modules
To prove Theorem 0.1, we need to pass from fermionic groups to . We will do this by the following chain of symmetric monoidal functors, whose notation we will explain and/or recall below:
(1) Recall from Definition 1.5 that is a full subcategory of . The functor composes the map with the -map from Equation (2.15). By Proposition 1.1, the -structure on induces a symmetric monoidal structure on .
The superscript means finite type: since we have restricted to finite fermionic groups, their classifying spaces are finite type.
(2) The functor is Ando–Blumberg–Gepner–Hopkins–Rezk’s Thom spectrum functor [1 Definition 2.20], which by Ando–Blumberg–Gepner [3 Theorem 1.6] is symmetric monoidal. Moreover, the Thom isomorphism theorem (specifically its incarnation in [47 Lemma 2.4]) can be used to show that sends finite-type spaces over to finite-type -modules.
(3) For a -module , let [EKMM97, Definition IV.1.7], where carries the --algebra structure given by Postnikov truncation followed by reduction mod 2. This is naturally a module over , which Baker [11 Theorem 5.1] shows has a unique isomorphism to . Like for ordinary cohomology, of a finite type -module is a finite-type -module, which means that for all , the subspace of homogeneous degree- elements is finite-dimensional. We will show below in Lemma 3.4 that is a symmetric monoidal functor via the cross product map .
(4) Finally, is the -localization functor from Theorem 2.7, which Ricka [83 Proposition 2.6] showed is symmetric monoidal. This needs no finiteness assumptions, so we implicitly compose with the forgetful functor from finite type -modules to all -modules, which is symmetric monoidal.
Thus is a symmetric monoidal functor of -categories, once we fulfill our promise about the cross product.
Definition 3.2. Let be the multiplication map in the --algebra structure on ko$-module maps and , define their cross product to be the composition
Passing to homotopy groups of mapping spectra, we get a map
which we also call the cross product.
Lemma 3.4. The functor is symmetric monoidal via the cross product . In particular, the cross product (34) is an isomorphism for all finite type -modules and .
The proof of Lemma 3.4 is identical to the usual proof of the standard fact that, by the Künneth formula, mod 2 cohomology is a symmetric monoidal functor out of the -category of finite type spaces. Absent the finite type hypothesis, and are only lax symmetric monoidal.
Lemma 3.5. In , and .
Proof. After unwinding the definition, is (the -localization of) of the Thom spectrum of the map , where is the generator. If denotes the tautological line bundle, then , so , as follows by combining [47 Example 1.6 and Lemma 1.8]. For any spectrum , there is a natural -module isomorphism [DKS26, Lemma 5.16], where the -module structure on the latter is the restriction of the usual -module structure. Finally, Campbell [25 Example 6.4] shows . As this is already -local by [Yu95, Chapter 2], . As we saw in Figure 1, is an invertible object of , and invertibility is closed under suspension (since ).
For , the story is very similar, but with in place of as the map to , in place of , and in place of . Campbell [25 Example 6.6] shows , and is -local [Yu95, Chapter 2], so the -localization of is itself, and it is an invertible object of , as displayed in Figure 1. ∎
Corollary 3.6. For all , is invertible.
Proof. By Lemma 3.5, and , and by Theorem 2.10, both of these are invertible in . Since is symmetric monoidal, is a tensor product of copies of and , and a tensor product of invertible objects is invertible.
Corollary 3.7. restricts to a homomorphism of commutative monoids
We will frequently compose with the isomorphism from Theorem 2.10.
Proofs of the main theorems
Our first main result, Theorem 0.1, is that the map from Corollary 3.7 is a group completion. That is, we will show that any map to a group factors uniquely through .
Proof of Theorem 0.1. Using the monoid isomorphism from Lemma 1.23, it suffices to show that the following map is a group completion:
Let be a group and be a monoid homomorphism. We may without loss of generality assume is abelian and write elements of additively. Define
Since by the relation in the domain of , only depends on mod , so indeed is well-defined as a homomorphism of abelian groups.
Since and , . To show this factorization is unique, suppose is another map with . Thus and . That is, and agree on a generating set of , so . □
Finally, we compare the two s that we introduced in §2.1 and §2.2, which gave rise to our two periodicities. We do this in two ways. First, in Proposition 4.3, we compute the “obvious” map , namely the map induced from by the universal property of the group completion. This map turns out to be —not an isomorphism. After discussing what went wrong, we prove Theorem 0.6, offering a different way to identify these two s as generated by the fermionic group .
Proposition 4.3. The map defined in (0.5) is multiplication by 2.
Proof. Recall that t:\mathbb{Z}/4\lhook\joinrel\longrightarrow\mathbb{Z}\oplus\mathbb{Z}/4 is the inclusion of the torsion subgroup and is the unique map such that , as guaranteed by Theorem 0.1. We calculate on the generator of , which is the module by Theorem 2.10: we have , so
By Example 1.30, and . As we saw in Example 2.32, these correspond to 1, resp. 3 in under Lemma 2.23, so we conclude that . □
One might have expected Proposition 4.3 to be an isomorphism. The issue stems from the different requirements for normalization between the Picard groupoid and the group of twists. In the former, torsion elements might not begin in degree zero, but in the latter, twists are given by rank zero virtual bundles. In particular, by Lemma 3.5, is not a torsion element of . Instead, it is the suspension of a torsion element. Generally, one does not think of and as vastly different modules. But this means that is not represented just by : now, the suspension matters, and we have to represent by the formal difference .
One would want to modify to fix this, but the setup is very rigid: there is a unique subgroup of , and the group completion is essentially unique. So, while does not reflect how suspension shouldn’t change the periodicity of -locally invertible -modules, there is not a good way to modify . This is why we formulate Theorem 0.6 as a correspondence.
Proof of Theorem 0.6. We compute each side of the diagram. On the left, we have by definition of . By Example 1.30, , and by Example 2.32, this is a generator of .
On the right, by definition suspends or desuspends a module until it is stably equivalent to , , , or (see Lemma 2.12). By Lemma 3.5, , so , which by Theorem 2.10 generates the torsion subgroup of . □
References
Department of Mathematics, The University of Kentucky, 719 Patterson Office Tower, Lexington, KY 40506, USA Email address: [email protected] URL: https://adebray.github.io/
Perimeter Institute for Theoretical Physics, 31 Caroline St North, Waterloo, ON, N2L 2Y5, Canada Email address: [email protected] URL: https://cakrulewski.github.io/
Department of Mathematics, University of California, Berkeley, 970 Evans Hall, Berkeley, CA 94720, USA Email address: [email protected] URL: https://natipt.github.io/
References
- [1]Matthew Ando, Andrew J. Blumberg, David Gepner, Michael J. Hopkins, and Charles Rezk. An ∞-categorical approach to R-line bundles, R-module Thom spectra, and twisted R-homology. J. Topol., 7(3):869–893, 2014. https://arxiv.org/abs/1403.4325. 12, 13, 16
- [2]Matthew Ando, Andrew J. Blumberg, David Gepner, Michael J. Hopkins, and Charles Rezk. Units of ring spectra, orientations and Thom spectra via rigid infinite loop space theory. J. Topol., 7(4):1077–1117, 2014. https://arxiv.org/abs/1403.4320. 12
- [3]Matthew Ando, Andrew J. Blumberg, and David Gepner. Parametrized spectra, multiplicative Thom spectra and the twisted Umkehr map. Geom. Topol., 22(7):3761–3825, 2018. https://arxiv.org/abs/1112.2203. 4, 13, 16
- [5]M. F. Atiyah, R. Bott, and A. Shapiro. Clifford modules. Topology, 3(suppl. suppl. 1):3–38, 1964. 8, 13
- [6]Benjamin Antieau, David Gepner, and José Manuel Gómez. Actions of K(π, n) spaces on K-theory and uniqueness of twisted K-theory. Trans. Amer. Math. Soc., 366(7):3631–3648, 2014. https://arxiv.org/abs/1106.5099. 13
- [7]J. F. Adams and H. R. Margolis. Modules over the Steenrod algebra. Topology, 10:271–282, 1971. 10DOI
- [8]Helena Albuquerque and Shahn Majid. Clifford algebras obtained by twisting of group algebras. Journal of Pure and Applied Algebra, 171(2-3):133–148, 2002. https://arxiv.org/abs/math/0011040. 8
- [9]Jan Albert, Masaki Okada, and Yuji Tachikawa. The art of orbifolding by huge non-abelian groups. 2026. https://arxiv.org/abs/2610.36024. 7
- [10]J. F. Adams and S. B. Priddy. Uniqueness of BSO. Math. Proc. Cambridge Philos. Soc., 80(3):475–509, 1976. 11, 14DOI
- [11]Andrew Baker. Homotopy theory of modules over a commutative S-algebra: some tools and examples. 2020. https://arxiv.org/abs/2003.12003. 15, 17
- [12]Dave Benson. Spin modules for symmetric groups. J. London Math. Soc. (2), 38(2):250–262, 1988. 3, 6, 8DOI
- [13]Robert R. Bruner and J. P. C. Greenlees. Connective real K-theory of finite groups, volume 169 of Mathematical Surveys and Monographs. American Mathematical Society, Providence, RI, 2010. 14
- [14]Paul Balmer and Martin Gallauer. The geometry of permutation modules. Invent. Math., 241(3):841–928, 2025. https://arxiv.org/abs/2307.04398. 15
- [15]Andrew Baker and Andrej Lazarev. On the Adams spectral sequence for R-modules. Algebr. Geom. Topol., 1:173–199, 2001. https://arxiv.org/abs/math/0105079. 15
- [16]Agnès Beaudry, Chloe Lewis, Clover May, Sabrina Pauli, and Elizabeth Tatum. A guide to equivariant parametrized cohomology. Topology Appl., 376:Paper No. 109449, 71, 2025. https://arxiv.org/abs/2410.13971. 15
- [17]Jonathan Beardsley, Kiran Luecke, and Jack Morava. Brauer–Wall groups and truncated Picard spectra of K-theory. Algebr. Geom. Topol., 26(5):1749–1780, 2026. https://arxiv.org/abs/2306.10112. 5, 6
- [19]J. Michael Boardman. Conditionally convergent spectral sequences. In Homotopy invariant algebraic structures (Baltimore, MD, 1998), volume 239 of Contemp. Math., pages 49–84. Amer. Math. Soc., Providence, RI, 1999. 15
- [20]Prasit Bhattacharya and Nicolas Ricka. The stable Picard group of A(2). 2017. https://arxiv.org/abs/1702.01493. 11
- [21]David Brook. Computations in higher twisted K-theory. Master’s thesis, The University of Adelaide, 2020. https://arxiv.org/abs/2007.08964. 14
- [22]Robert R. Bruner. Idempotents, localizations and Picard groups of A(1)-modules. In An alpine expedition through algebraic topology, volume 617 of Contemp. Math., pages 81–108. Amer. Math. Soc., Providence, RI, 2014. https://arxiv.org/abs/1211.0213. 1, 11
- [23]Thomas Brazelton and Matthias Wendt. The Chow–Witt rings of the classifying space of quadratically oriented bundles. 2025. https://arxiv.org/abs/2501.06302. 13
- [24]Martin Čadek. The cohomology of BO(n) with twisted integer coefficients. J. Math. Kyoto Univ., 39(2):277–286, 1999. 13DOI
- [25]Jonathan A. Campbell. Homotopy theoretic classification of symmetry protected phases. 2017. https://arxiv.org/abs/1708.04264. 16, 17
- [34]J. H. Conway. A perfect group of order 8, 315, 553, 613, 086, 720, 000 and the sporadic simple groups. Proc. Nat. Acad. Sci. U.S.A., 61:398–400, 1968. 6pubmed.ncbi.nlm.nih.gov/16591697
- [35]J. H. Conway. A group of order 8, 315, 553, 613, 086, 720, 000. Bull. London Math. Soc., 1:79–88, 1969. 6DOI
- [47]Arun Debray and Matthew Yu. Adams spectral sequences for non-vector-bundle Thom spectra. 2023. https://arxiv.org/abs/2305.01678. 5, 6, 14, 15, 16, 17
- [55]Juan Omar Gómez and Sam K. Miller. The classification of integral endotrivial complexes. 2026. https://arxiv.org/abs/2605.31128. 15
- [56]Juan Omar Gómez and Sam K. Miller. Permutation twisted cohomology, reduced and immersed. 2026. https://arxiv.org/abs/2610.40602. 15
- [60]Michael J. Hopkins, Mark Mahowald, and Hal Sadofsky. Constructions of elements in Picard groups. In Topology and representation theory (Evanston, IL, 1992), volume 158 of Contemp. Math., pages 89–126. Amer. Math. Soc., Providence, RI, 1994.DOI
- [61]Takumi Ichikawa and Yuji Tachikawa. The super Frobenius-Schur indicator and finite group gauge theories on pin− surfaces. Comm. Math. Phys., 400(1):417–428, 2023. https://arxiv.org/abs/2002.10642.
- [63]Michael Joachim. Higher coherences for equivariant K-theory. In Structured ring spectra, volume 315 of London Math. Soc. Lecture Note Ser., pages 87–114. Cambridge Univ. Press, Cambridge, 2004.DOI
- [64]Max Karoubi. Algèbres de Clifford et K-théorie. Ann. Sci. École Norm. Sup. (4), 1:161–270, 1968.DOI
- [65]Alexei Kitaev. Periodic table for topological insulators and superconductors. In AIP Conference Proceedings, pages 22–30, 2009. https://arxiv.org/abs/0901.2686.
- [66]Matthias Kreck. Surgery and duality. Ann. of Math. (2), 149(3):707–754, 1999. https://arxiv.org/abs/math/9905211.
- [67]Anton Kapustin, Ryan Thorngren, Alex Turzillo, and Zitao Wang. Fermionic symmetry protected topological phases and cobordisms. J. High Energy Phys., (12):052, front matter+20pp, 2015. https://arxiv.org/abs/1406.7329.
- [69]Richard Lastovecki. Cohomology of BO(n₁) × · · · × BO(nₘ) with local integer coefficients. Comment. Math. Univ. Carolin., 46(1):21–32, 2005.
- [70]Jacob Lurie. Higher algebra. 2017. https://www.math.ias.edu/~lurie/papers/HA.pdf.
- [71]Ying-Hsuan Lin and Mayuko Yamashita. Topological elliptic genera I – the mathematical foundation. 2024. https://arxiv.org/abs/2412.02298.
- [72]H. R. Margolis. Spectra and the Steenrod algebra, volume 29 of North-Holland Mathematical Library. North-Holland Publishing Co., Amsterdam, 1983. Modules over the Steenrod algebra and the stable homotopy category.
- [73]J. Peter May. E∞ ring spaces and E∞ ring spectra. Lecture Notes in Mathematics, Vol. 577. Springer-Verlag, Berlin-New York, 1977. With contributions by Frank Quinn, Nigel Ray, and Jørgen Tornehave. https://www.math.uchicago.edu/~may/BOOKS/e_infty.pdf.DOI
- [74]J. P. May. Picard groups, Grothendieck rings, and Burnside rings of categories. Adv. Math., 163(1):1–16, 2001. https://www.math.uchicago.edu/~may/PAPERS/PicJan01.pdf.DOI
- [75]John Milnor. The Steenrod algebra and its dual. Ann. of Math. (2), 67:150–171, 1958.
- [76]Sam K. Miller. Permutation twisted cohomology, remixed. 2025. https://arxiv.org/abs/2509.00954.
- [77]I. Madsen, V. Snaith, and J. Tornehave. Infinite loop maps in geometric topology. Math. Proc. Cambridge Philos. Soc., 81(3):399–430, 1977.DOI
- [78]Ulrich Pennig. A non-commutative model for higher twisted K-theory. J. Topol., 9(1):27–50, 2016. https://arxiv.org/abs/1502.02807.
- [79]F. P. Peterson. Lectures on Cobordism Theory. Lectures in Mathematics. Kinokuniya Book Store Co., Ltd., 1968.
- [80]JianZhong Pan and RuJia Yan. The stable Picard group of A(n). J. Homotopy Relat. Struct., 21(1):23–43, 2026. https://arxiv.org/abs/2212.09985.
- [81]Raquel Queiroz, Eslam Khalaf, and Ady Stern. Dimensional hierarchy of fermionic interacting topological phases. Physical Review Letters, 117(20):206405, November 2016. https://arxiv.org/abs/1601.01596.
- [82]Andrew Ranicki. Eight in algebra, topology and mathematical physics. https://webhomes.maths.ed.ac.uk/~v1ranick/eight.htm.
- [83]Nicolas Ricka. Local study of stable module categories via tensor triangulated geometry. October 2016. https://arxiv.org/abs/1610.03561.
- [84]Luuk Stehouwer. Interacting SPT phases are not Morita invariant. Letters in Mathematical Physics, 112(3):64, June 2022. https://arxiv.org/abs/2110.07408.
- [85]Stephan Stolz. Exotic structures on 4-manifolds detected by spectral invariants. Invent. Math., 94(1):147–162, 1988. 16DOI
- [86]Stephan Stolz. Concordance classes of positive scalar curvature metrics, 1998. https://www3.nd.edu/~stolz/preprint.html. 3
- [87]Guo Chuan Thiang. On the K-theoretic classification of topological phases of matter. Annales Henri Poincaré, 17(4):757–794, May 2015. https://arxiv.org/abs/1406.7366. 8
- [88]Yuji Tachikawa and Kazuya Yonekura. Why are fractional charges of orientifolds compatible with Dirac quantization? SciPost Phys., 7:58, 2019. https://arxiv.org/abs/1805.02772. 1, 16
- [89]Jeroen van der Meer and Richard Wong. Endotrivial modules for cyclic p-groups and generalized quaternion groups via Galois descent. 2021. https://arxiv.org/abs/2107.06308. 11
- [90]C. T. C. Wall. A characterization of simple modules over the Steenrod algebra mod 2. Topology, 1:249–254, 1962. 10DOI
- [91]Bai-Ling Wang. Geometric cycles, index theory and twisted K-homology. J. Noncommut. Geom., 2(4):497–552, 2008. https://arxiv.org/abs/0710.1625. 15
- [92]Matthias Wendt. Chow-Witt rings of Grassmannians. Algebr. Geom. Topol., 24(1):1–48, 2024. https://arxiv.org/abs/1805.06142. 13
- [93]Zheyan Wan, Juven Wang, and Yunqin Zheng. Higher anomalies, higher symmetries, and cobordisms II: Lorentz symmetry extension and enriched bosonic/fermionic quantum gauge theory. Ann. Math. Sci. Appl., 5(2):171–257, 2020. https://arxiv.org/abs/1912.13504. 1, 16
- [94]Lucy Yang. Categorical dynamics on Stable Module Categories. PhD thesis, Harvard University, 2022. https://arxiv.org/abs/2212.09964. 10
