A categorical formulation of the Deligne-Terasoma approach to double shuffle theory
Benjamin Enriquez, Khalef Yaddaden · 2025
Read A categorical formulation of the Deligne-Terasoma approach to double shuffle theory by Benjamin Enriquez, Khalef Yaddaden (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-15348.txt.
Introduction
Benjamin Enriquez and Khalef Yaddaden recast a deep algebra theory in category terms in 2025. Double shuffle theory links two ways of multiplying special values in math. A geometric approach by Deligne and Terasoma frames those links through harmonic products. The authors seek a cleaner base for that construction.
They define a bimodule that carries a factorization structure. Such a structure is shown to yield an algebra map on its own. Both harmonic products of the theory then fall out of the frame. A categorical base may make a hard theory easier to extend and teach.
Catalog record
| CREATOR | Benjamin Enriquez, Khalef Yaddaden |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 181 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-15348
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A categorical formulation of the Deligne-Terasoma approach to double shuffle theory Benjamin Enriquez, Khalef Yaddaden Abstract . In this paper, we introduce the notion of a bimodule with a factorization structure (BFS) and show that such a structure gives rise to an algebra morphism. We then prove that this framework offers an interpretation of the geometric construction underlying both the Betti and de Rham harmonic coproducts of the double shuffle theory developed in [ DeT , EF1 , EF2 , EF3 ] . 2020 Mathematics Subject Classification Primary 11M32, 16D20. Secondary 16W70, 20F36. Introduction In [ EF1 ] , the first author and Furusho revisited the formalism of double shuffle relations among multiple zeta values set up by Racinet [ Rac ] . There, they explained the “de Rham” nature of this formalism and constructed its “Betti” version, whose main objects are algebra “harmonic” coproducts \Delta^{\mathcal{W},\mathrm{DR}} and \Delta^{\mathcal{W},\mathrm{B}} , respectively equipping the algebras \mathcal{W}^{\mathrm{DR}} and \mathcal{W}^{\mathrm{B}} with a Hopf algebra structure. The algebras \mathcal{W}^{\mathrm{DR}} and \mathcal{W}^{\mathrm{B}} can be seen as subalgebras of respectively \mathcal{V}^{\mathrm{B}} , the group algebra of the free group with two generators and \mathcal{V}^{\mathrm{DR}} , the enveloping algebra of the free Lie algebra with two generators. The authors used this formalism to prove that the associator relations between the multiple zeta values imply the double shuffle relations. This is formulated as the inclusion of the torsor of associators in the double shuffle torsor [ EF2 ] . To this end, they constructed isomorphisms relating the Betti and de Rham sides and showed that any associator relates the algebra coproducts \Delta^{\mathcal{W},\mathrm{DR}} and \Delta^{\mathcal{W},\mathrm{B}} ( [ EF1 , Theorem 10.9] ). Rather than relying on Bar constructions as in [ Fur ] , this alternative proof builds upon an interpretation of the harmonic coproducts in terms of infinitesimal braid Lie algebras for the de Rham side and braid groups for the Betti side, which is implicit in the unpublished work of Deligne and Terasoma [ DeT ] . The purpose of this paper is to formulate this construction of the harmonic coproducts in a categorical framework, which will be used in a later paper [ EY2 ] for the geometric interpretation of the cyclotomic version of the Betti and de Rham coproducts introduced by the second author in [ Yad ] . To this end, we define the category \mathsf{BFS} of bimodules with factorization structures whose objects are tuples (\mathbf{B},\mathbf{M},\mathbf{A},\rho,e,\boldsymbol{r},\boldsymbol{c}) such that (\mathbf{B},\mathbf{M},\mathbf{A},\rho) is a bimodule (i.e. \mathbf{M} is a right \mathbf{A} -module equipped with a compatible left \mathbf{B} -action \rho ) and (e,\boldsymbol{r},\boldsymbol{c}) satisfies the factorization identity (see Definition 2.1.1 ) \rho(e)=\boldsymbol{c}\circ\boldsymbol{r}. We then construct a functor \mathsf{BFS}\to\mathsf{Mor}(\mathsf{alg}) which associates to any object (\mathbf{B},\mathbf{M},\mathbf{A},\rho,e,\boldsymbol{r},\boldsymbol{c}) of \mathsf{BFS} , an algebra morphism given by (see Proposition-Definition 2.3.2 ) (0.1) \mathbf{B}\ni b\mapsto\boldsymbol{r}\circ\rho(b)\circ\boldsymbol{c}(1_{\mathbf{A}})\in\mathbf{A}. On the other hand, we also define categories \mathsf{BFS}_{\mathrm{fil}} and \mathsf{BFS}_{\mathrm{gr}} of respectively filtered and graded objects of \mathsf{BFS} as well as a functor \mathsf{BFS}_{\mathrm{fil}}\to\mathsf{BFS}_{\mathrm{gr}} which is induced by associated graded objects. This enables the construction of functors \mathsf{BFS}_{\mathrm{fil}}\to\mathsf{Mor}(\mathsf{alg}_{\mathrm{fil}}) and \mathsf{BFS}_{\mathrm{gr}}\to\mathsf{Mor}(\mathsf{alg}_{\mathrm{gr}}) fitting in the following diagram (0.2) {\lx@inpgf@ignorespaces\mathsf{BFS}} {\lx@inpgf@ignorespaces\mathsf{BFS}_{\mathrm{fil}}} {\lx@inpgf@ignorespaces\mathsf{BFS}_{\mathrm{gr}}} {\lx@inpgf@ignorespaces\mathsf{Mor}(\mathsf{alg})} {\lx@inpgf@ignorespaces\mathsf{Mor}(\mathsf{alg}_{\mathrm{fil}})} {\lx@inpgf@ignorespaces\mathsf{Mor}(\mathsf{alg}_{\mathrm{gr}})} Sec. 2.3.1 ( 2.5 ) Sec. 2.2 Sec. 2.3.2 Sec. 2.3.3 ( 2.8 ) ( 2.7 ) Each of the squares gives rise to two functors, and one checks that there are natural equivalences relating them. The Betti and de Rham formalism of [ EF1 ] can be interpreted within this framework through the objects (see Proposition-Definitions 4.7.1 and 4.7.3 ) \mathcal{O}^{\mathrm{B}}:=(\mathcal{V}^{\mathrm{B}},\mathbf{M}^{\mathrm{B}},\mathcal{V}^{\mathrm{B}}\otimes\mathcal{V}^{\mathrm{B}},\rho^{\mathrm{B}},e^{\mathrm{B}},\boldsymbol{r}^{\mathrm{B}},\boldsymbol{c}^{\mathrm{B}})\in\mathsf{BFS}, and \mathcal{O}^{\mathrm{DR}}:=(\mathcal{V}^{\mathrm{DR}},\mathbf{M}^{\mathrm{DR}},\mathcal{V}^{\mathrm{DR}}\otimes\mathcal{V}^{\mathrm{DR}},\rho^{\mathrm{DR}},e^{\mathrm{DR}},\boldsymbol{r}^{\mathrm{DR}},\boldsymbol{c}^{\mathrm{DR}})\in\mathsf{BFS}_{\mathrm{gr}}. A suitable filtration on the object \mathcal{O}^{\mathrm{B}} enables us to show that it gives rise to an object \mathcal{O}^{\mathrm{B}}_{\mathrm{fil}}\in\mathsf{BFS}_{\mathrm{fil}} (see Corollary 4.7.2 ). Thanks to the functor \mathsf{BFS}_{\mathrm{fil}}\to\mathsf{BFS}_{\mathrm{gr}} these objects are related by the following: Theorem I (Theorem 4.6.4 ) . We have \mathrm{gr}(\mathcal{O}^{\mathrm{B}}_{\mathrm{fil}})=\mathcal{O}^{\mathrm{DR}} . The bimodule parts of these objects arise from geometric constructions, but their factorization structures (e^{\mathrm{B}},\boldsymbol{r}^{\mathrm{B}},\boldsymbol{c}^{\mathrm{B}}) and (e^{\mathrm{DR}},\boldsymbol{r}^{\mathrm{DR}},\boldsymbol{c}^{\mathrm{DR}}) require the use of explicit objects of the BFS categories arising from representations of the algebras \mathcal{V}^{\mathrm{B}} and \mathcal{V}^{\mathrm{DR}} , namely (see Corollaries 3.1.4 and 3.4.5 ) \mathcal{O}^{\mathrm{B}}_{\mathsf{mat}}:=(\mathcal{V}^{\mathrm{B}},(\mathcal{V}^{\mathrm{B}}\otimes\mathcal{V}^{\mathrm{B}})^{\oplus 3},\mathcal{V}^{\mathrm{B}}\otimes\mathcal{V}^{\mathrm{B}},\mathsf{r}\underline{\rho},e^{\mathrm{B}},\mathsf{r}\underline{\mathrm{row}},\mathsf{r}\underline{\mathrm{col}})\in\mathsf{BFS}, and \mathcal{O}^{\mathrm{DR}}_{\mathsf{mat}}:=(\mathcal{V}^{\mathrm{DR}},(\mathcal{V}^{\mathrm{DR}}\otimes\mathcal{V}^{\mathrm{DR}})^{\oplus 3},\mathcal{V}^{\mathrm{DR}}\otimes\mathcal{V}^{\mathrm{DR}},\mathsf{r}\rho,e^{\mathrm{DR}},\mathsf{r}\mathrm{row},\mathsf{r}\mathrm{col})\in\mathsf{BFS}_{\mathrm{gr}}. The prefix “ \mathsf{r} ” signifies that the constructed objects differ from the objects with the same notation from [ EF1 ] by the fact that we consider right actions instead of left actions. Then, once again, a suitable filtration on the object \mathcal{O}^{\mathrm{B}}_{\mathsf{mat}} enables us to show that it gives rise to an object \mathcal{O}^{\mathrm{B}}_{\mathsf{mat},\mathrm{fil}}\in\mathsf{BFS}_{\mathrm{fil}} (see Corollary 3.3.4 ). We show that the geometric objects \mathcal{O}^{\mathrm{B}}_{\mathrm{fil}} and \mathcal{O}^{\mathrm{DR}} and the explicit objects \mathcal{O}^{\mathrm{B}}_{\mathsf{mat},\mathrm{fil}} and \mathcal{O}^{\mathrm{DR}}_{\mathsf{mat}} are respectively related thanks to bimodule isomorphisms: Theorem II (Theorems 4.3.4 and 4.5.4 ) . We have \mathbf{M}^{\mathrm{B}}\simeq(\mathcal{V}^{\mathrm{B}}\otimes\mathcal{V}^{\mathrm{B}})^{\oplus 3}\text{ and }\mathbf{M}^{\mathrm{DR}}\simeq(\mathcal{V}^{\mathrm{DR}}\otimes\mathcal{V}^{\mathrm{DR}})^{\oplus 3}. Finally, the explicit BFS objects enables the use of the functor \mathsf{BFS}\to\mathsf{Mor}(\mathsf{alg}) to construct coproducts \Delta_{\mathcal{O}^{\mathrm{B}}_{\mathsf{mat}}} and \Delta_{\mathcal{O}^{\mathrm{DR}}_{\mathsf{mat}}} , as described in ( 0.1 ); then we identify them with the coproducts \Delta^{\mathcal{W},\mathrm{B}} and \Delta^{\mathcal{W},\mathrm{DR}} thanks to the following result: Theorem III (Theorems 3.2.1 and 3.5.1 ) . For b in \mathcal{V}^{\mathrm{B}} (resp. \mathcal{V}^{\mathrm{DR}} ), we have \Delta_{\mathcal{O}^{\mathrm{B}}_{\mathsf{mat},\mathrm{fil}}}(b)=\Delta^{\mathcal{W},\mathrm{B}}(b\ e^{\mathrm{B}})\quad(\text{resp.}\quad\Delta_{\mathcal{O}^{\mathrm{DR}}_{\mathsf{mat}}}(b)=\Delta^{\mathcal{W},\mathrm{DR}}(b\ e^{\mathrm{DR}})). Acknowledgments . This project was partially supported by first author’s ANR grant Project HighAGT ANR20-CE40-0016 and second author’s JSPS KAKENHI Grant 23KF0230. Notation . Throughout this paper, let \mathbf{k} be a commutative \mathbb{Q} -algebra. 1. Basic categories of filtered and graded algebras and modules We introduce here the basic categories and functors relating them that we shall refer to throughout this paper. 1.1. The categories \mathbf{k}{\text{-}}\mathsf{mod} , \mathbf{k}{\text{-}}\mathsf{mod}_{\mathrm{fil}} and \mathbf{k}{\text{-}}\mathsf{mod}_{\mathrm{gr}} Definition 1.1.1 . (a) \mathbf{k}\text{-}\mathsf{mod} is the category of \mathbf{k} -modules; (b) \mathbf{k}{\text{-}}\mathsf{mod}_{\mathrm{fil}} is the category of filtered \mathbf{k} -modules, that is, \mathbf{k} -modules \mathbf{M} equipped with a decreasing sequence of \mathbf{k} -submodules (\mathcal{F}^{n}\mathbf{M})_{n\in\mathbb{Z}} called filtration. Morphisms are filtered \mathbf{k} -module morphisms, that is, \mathbf{k} -module morphisms \boldsymbol{\varphi}:\mathbf{M}\to\mathbf{M}^{\prime} which are compatible with the filtrations on both sides. We denote by \mathcal{F}^{n}\boldsymbol{\varphi}:\mathcal{F}^{n}\mathbf{M}\to\mathcal{F}^{n}\mathbf{M}^{\prime} the induced \mathbf{k} -module morphism corresponding to n\in\mathbb{Z} ; (c) \mathbf{k}{\text{-}}\mathsf{mod}_{\mathrm{gr}} is the category of \mathbb{Z} - graded \mathbf{k} -modules, that is, \mathbf{k} -module for which there exists a sequence of \mathbf{k} -submodules (\mathbf{M}_{n})_{n\in\mathbb{Z}} , called grading, such that \mathbf{M}=\bigoplus_{n\in\mathbb{Z}}\mathbf{M}_{n} . Morphisms are graded \mathbf{k} -module morphisms, that is, \mathbf{k} -module morphisms \boldsymbol{\varphi}:\mathbf{M}\to\mathbf{M}^{\prime} such that \varphi(\mathbf{M}_{n})\subset\mathbf{M}_{n}^{\prime} for any n\in\mathbb{Z} . We denote by \boldsymbol{\varphi}_{n}:\mathbf{M}_{n}\to\mathbf{M}^{\prime}_{n} the induced \mathbf{k} -module morphism corresponding to n\in\mathbb{Z} . Recall that there is an associated graded functor \mathrm{gr}:\mathbf{k}{\text{-}}\mathsf{mod}_{\mathrm{fil}}\rightarrow\mathbf{k}{\text{-}}\mathsf{mod}_{\mathrm{gr}} which takes a filtered \mathbf{k} -module \left(\mathbf{M},(\mathcal{F}^{n}\mathbf{M})_{n\in\mathbb{Z}}\right) to the graded \mathbf{k} -module \mathrm{gr}(\mathbf{M}):=\bigoplus_{n\in\mathbb{Z}}\mathrm{gr}_{n}(\mathbf{M}) , where \mathrm{gr}_{n}(\mathbf{M}):=\mathcal{F}^{n}\mathbf{M}/\mathcal{F}^{n+1}\mathbf{M} for any n\in\mathbb{Z} . Denote by x\mapsto[x]_{n} the canonical projection \mathcal{F}^{n}\mathbf{M}\twoheadrightarrow\mathrm{gr}_{n}\mathbf{M} for any n\in\mathbb{Z} . At the level of morphisms, one assigns to a filtered \mathbf{k} -module morphism \boldsymbol{\varphi}:\mathbf{M}\to\mathbf{M}^{\prime} , the graded \mathbf{k} -module morphism \mathrm{gr}(\boldsymbol{\varphi}):\mathrm{gr}(\mathbf{M})\to\mathrm{gr}(\mathbf{M}^{\prime}) induced by the commutative diagram {\lx@inpgf@ignorespaces\mathcal{F}^{n}\mathbf{M}} {\lx@inpgf@ignorespaces\mathcal{F}^{n}\mathbf{M}^{\prime}} {\lx@inpgf@ignorespaces\mathrm{gr}_{n}(\mathbf{M})} {\lx@inpgf@ignorespaces\mathrm{gr}_{n}(\mathbf{M}^{\prime})} \scriptstyle{\lx@inpgf@ignorespaces\mathcal{F}^{n}\boldsymbol{\varphi}} for any n\in\mathbb{Z} (see for example [ Bbk , Chap. III, Sec. 2, no. 3 and no. 4] ). 1.2. The categories \mathbf{k}{\text{-}}\mathsf{alg} , \mathbf{k}{\text{-}}\mathsf{alg}_{\mathrm{fil}} and \mathbf{k}{\text{-}}\mathsf{alg}_{\mathrm{gr}} Definition 1.2.1 . (a) \mathbf{k}\text{-}\mathsf{alg} is the category of \mathbf{k} -algebras (with unit); (b) \mathbf{k}{\text{-}}\mathsf{alg}_{\mathrm{fil}} is the category of filtered \mathbf{k} -algebras, that is, \mathbf{k} -algebras (\mathbf{A},\cdot) equipped with a \mathbf{k} -module filtration (\mathcal{F}^{n}\mathbf{A})_{n\in\mathbb{Z}} that satisfies \mathcal{F}^{k}\mathbf{A}\cdot\mathcal{F}^{n}\mathbf{A}\subset\mathcal{F}^{k+n}\mathbf{A} , for any k,n\in\mathbb{Z} and 1\in\mathcal{F}^{0}\mathbf{A} . Morphisms are filtered \mathbf{k} -algebra morphisms, that is, \mathbf{k} -algebra morphisms which are also filtered \mathbf{k} -module morphisms; (c) \mathbf{k}{\text{-}}\mathsf{alg}_{\mathrm{gr}} is the category of