--- title: Arithmetic Dijkgraaf–Witten Theory url: https://www.emergentmind.com/topics/arithmetic-dijkgraaf-witten-theory type: topic --- # Arithmetic Dijkgraaf–Witten Theory Arithmetic Dijkgraaf–Witten theory is the arithmetic analogue of finite-group Dijkgraaf–Witten topological quantum field theory, obtained by replacing closed oriented \(3\)-manifolds with arithmetic curves such as \(X=\operatorname{Spec}(\mathcal{O}_F)\) or \(X=\operatorname{Spec}(\mathcal{O}_F)\setminus S\), replacing flat \(G\)-bundles by continuous Galois representations, and replacing evaluation on a fundamental class by étale or Galois cohomological invariant maps such as \(\operatorname{inv}:H^3(X,\mu_n)\simeq \tfrac{1}{n}\mathbb{Z}/\mathbb{Z}\) or \(\operatorname{inv}:H^3(X,\mathbf{G}_m)\simeq \mathbb{Q}/\mathbb{Z}\) [1510.05818]. In the literature this theme appears in several forms: Kim’s arithmetic Chern–Simons functionals on spaces of Galois representations [1510.05818], their boundary and \(p\)-adic refinements [1609.03012], Hirano’s extension to arbitrary number fields via modified étale cohomology [1911.12964], explicit mod \(2\) formulas for real quadratic fields [2308.05467], a pro-\(p\) cobordism-theoretic arithmetic field theory with an arithmetic Dijkgraaf–Witten example [2504.19078], and a duality theory parallel to Naidu’s classification of equivalent topological Dijkgraaf–Witten theories [2508.19214]. ## 1. Arithmetic replacement of the finite-group Dijkgraaf–Witten datum In ordinary \(3\)-dimensional Dijkgraaf–Witten theory one fixes a finite group \(G\) and a \(3\)-cocycle \[ \omega \in Z^3(G,U(1)) \quad\text{or equivalently } c \in Z^3(G,\mathbb{Q}/\mathbb{Z}), \] and for a closed oriented \(3\)-manifold \(M\) a classical gauge field is a flat \(G\)-bundle, equivalently a homomorphism \[ \rho:\pi_1(M)\to G \] up to conjugacy. The partition function is \[ Z_\omega(M)=\sum_{[\rho:\pi_1(M)\to G]} \frac{1}{|\operatorname{Aut}(\rho)|}\exp(2\pi i\,S_\omega(\rho)), \] with \(S_\omega(\rho)\) obtained by pulling back \(\omega\) and evaluating on the fundamental class of \(M\) [1609.03012]. The arithmetic analogue replaces \(M\) by \[ X=\operatorname{Spec}(\mathcal{O}_F), \] where \(F\) is a number field, often totally imaginary in the initial formulations. The étale fundamental group \[ \pi:=\pi_1(X,b) \] plays the role of \(\pi_1(M)\), and the cohomological input is the canonical isomorphism \[ \operatorname{inv}: H^3(X,\mathbf{G}_m) \;\simeq\; \mathbb{Q}/\mathbb{Z}, \] together with the finite-coefficient variants \[ \operatorname{inv}: H^3\bigl(X,\mu_n\bigr) \;\simeq\; \tfrac{1}{n}\mathbb{Z}/\mathbb{Z} \] and \[ \operatorname{inv}: H^3\bigl(X,\mathbb{Z}_p(1)\bigr) \;\simeq\; \mathbb{Z}_p. \] These are interpreted as arithmetic analogues of evaluation on the fundamental class of a compact oriented \(3\)-manifold [1609.03012]. For a finite group \(A\), the arithmetic configuration space is the moduli set \[ M(A):=\operatorname{Hom}_{\mathrm{cont}}(\pi,A)/A, \] the set of isomorphism classes of principal \(A\)-bundles in the étale topology. This is the arithmetic counterpart of the groupoid of flat \(A\)-bundles in topological Dijkgraaf–Witten theory [1609.03012]. A decisive extension beyond totally imaginary fields is Hirano’s use of the Artin–Verdier site \(X_{\mathrm{et}}\), the modified étale fundamental group \[ \pi_1(X):=\pi_1^{\text{Artin–Verdier}}(X), \] and modified étale cohomology \(H^i(X,\mathbb{Z}/n\mathbb{Z})\), so that arbitrary number fields, including those with real places, enter the theory through a \(3\)-dimensional duality framework analogous to Poincaré duality [1911.12964]. ## 2. Arithmetic Chern–Simons functionals and the boundary formalism Kim’s foundational construction begins with a finite group \(A\), a class \[ c\in H^3(A,\mathbb{Z}/n\mathbb{Z}), \] a totally imaginary number field \(F\) containing \(\mu_n\), and a trivialization \[ \zeta : \mathbb{Z}/n\mathbb{Z} \xrightarrow{\sim} \mu_n. \] For \([\rho]\in M(A)\), the arithmetic Chern–Simons functional is \[ \mathrm{CS}_c([\rho]) := \operatorname{inv}\bigl(\rho^{*}c\bigr)\in \tfrac{1}{n}\mathbb{Z}/\mathbb{Z}. \] This is the direct arithmetic analogue of the Dijkgraaf–Witten action obtained by pulling back a group \(3\)-cocycle and integrating it against a \(3\)-dimensional fundamental class [1510.05818]. When one removes a finite set of primes \(S\), the open arithmetic curve \[ X_S:=\operatorname{Spec}(\mathcal{O}_F[1/S])=X\setminus\{P_v:v\in S\} \] plays the role of a \(3\)-manifold with boundary. Its arithmetic fundamental group is \[ \pi_S:=\pi_1(X_S,b), \] and for each \(v\in S\) there is a local Galois group \[ \pi_v:=\operatorname{Gal}(F_v^{\mathrm{sep}}/F_v) \] with maps \(i_v:\pi_v\to\pi_S\). The local and global representation groupoids are \[ Y_S(A):=\operatorname{Hom}_{\mathrm{cont}}(\pi_S,A),\qquad Y_S^{\mathrm{loc}}(A):=\prod_{v\in S}\operatorname{Hom}_{\mathrm{cont}}(\pi_v,A), \] together with the restriction functor \[ r_S:M_S(A)\to M_S^{\mathrm{loc}}(A). \] This is the arithmetic replacement of restricting a flat bundle to boundary components [1609.03012]. The boundary theory is organized by an arithmetic Chern–Simons line. For local boundary data \[ \rho_S=(\rho_v)_{v\in S}\in Y_S^{\mathrm{loc}}(A), \] one defines \[ C_S:=\prod_{v\in S} C^2(\pi_v,\mathbb{Z}/n\mathbb{Z}),\quad Z_S:=\prod_{v\in S} Z^2(\pi_v,\mathbb{Z}/n\mathbb{Z}),\quad B_S:=\prod_{v\in S} B^2(\pi_v,\mathbb{Z}/n\mathbb{Z}), \] then \[ H(\rho_S):=d^{-1}(c\circ \rho_S)/B_S, \] a torsor under \[ H_S:=\prod_{v\in S} H^2(\pi_v,\mathbb{Z}/n\mathbb{Z}) \simeq \prod_{v\in S}\tfrac{1}{n}\mathbb{Z}/\mathbb{Z}. \] Pushing out by \[ \Sigma:\prod_{v\in S}\tfrac{1}{n}\mathbb{Z}/\mathbb{Z}\to \tfrac{1}{n}\mathbb{Z}/\mathbb{Z} \] gives the arithmetic Chern–Simons line \[ L(\rho_S):=\Sigma_*H(\rho_S), \] a \(\tfrac{1}{n}\mathbb{Z}/\mathbb{Z}\)-torsor. The global Chern–Simons action with boundary becomes a section \[ \mathrm{CS}_c \in \Gamma\bigl(M_S(A),L\bigr), \] rather than merely a scalar-valued function [1609.03012]. A \(p\)-adic version replaces the finite group \(A\) by a \(p\)-adic Lie group with an open surjection \[ t:A\to \Gamma:=\mathbb{Z}_p, \] takes \[ c\in H^3\bigl(A,\mathbb{Z}_p[[\Gamma]]\bigr), \] and uses a \(\pi_S\)-equivariant isomorphism \[ \zeta_\rho : \mathbb{Z}_p[[\Gamma]]_\rho \xrightarrow{\sim} \Lambda:=\mathbb{Z}_p[[\Gamma]]\otimes \mathbb{Z}_p(1) \] to define a \(p\)-adic Chern–Simons invariant \[ \mathrm{CS}_c([\rho]) \in L_{\mathrm{glob}([\rho],\mathbb{Z}_p)}. \] The subsequent quantization is explicitly deferred; the papers focus on the classical action-functional side [1609.03012]. ## 3. General number fields and explicit mod \(2\) arithmetic Dijkgraaf–Witten invariants For arbitrary number fields \(K\), Hirano defines the arithmetic Chern–Simons functional using modified étale cohomology on the Artin–Verdier site. With \[ X=\operatorname{Spec}\mathcal{O}_K,\qquad A\ \text{finite},\qquad c\in H^3(A,\mathbb{Z}/n\mathbb{Z}), \] and \[ M(X,A):=\operatorname{Hom}_{\mathrm{cont}}(\pi_1(X),A)/A, \] the arithmetic Chern–Simons invariant of \(\rho\in M(X,A)\) is the image of \(c\) under \[ H^3(A,\mathbb{Z}/n\mathbb{Z}) \xrightarrow{\rho^*} H^3\bigl(\pi_1(X),\mathbb{Z}/n\mathbb{Z}\bigr) \xrightarrow{j_3} H^3(X,\mathbb{Z}/n\mathbb{Z}) \cong \mathbb{Z}/n\mathbb{Z}, \] where \(j_3\) is the edge map of the modified Hochschild–Serre spectral sequence \[ H^p\big(\pi_1(X), H^q(X,\mathbb{Z}/n\mathbb{Z})\big) \;\Rightarrow\; H^{p+q}(X,\mathbb{Z}/n\mathbb{Z}). \] The arithmetic Dijkgraaf–Witten invariant is then \[ Z_c(X) \;:=\; \sum_{\rho\in M(X,A)} \exp\left(\frac{2\pi i}{n}\,\operatorname{CS}_c(\rho)\right) \] [1911.12964]. In the basic case \[ A=\mathbb{Z}/n\mathbb{Z},\qquad c:=\mathrm{id}\cup\beta(\mathrm{id})\in H^3(A,\mathbb{Z}/n\mathbb{Z}), \] with \(\beta\) the Bockstein from \[ 0\to \mathbb{Z}/n\mathbb{Z}\xrightarrow{\times n}\mathbb{Z}/n^2\mathbb{Z}\to \mathbb{Z}/n\mathbb{Z}\to 0, \] Hirano proves that for a surjective \[ \rho:\pi_1(X)\to \mathbb{Z}/n\mathbb{Z}, \] if \[ L=K(v^{1/n}) \] is the corresponding unramified Kummer extension and \(a\in I_K\) satisfies \[ a^n=(v)^{-1}, \] then \[ \operatorname{CS}_c(\rho)=\chi\big((L/K/a)\big), \] where \((L/K/a)\) is the Artin symbol and \[ \chi:\operatorname{Gal}(L/K)\xrightarrow{\sim}\mathbb{Z}/n\mathbb{Z} \] is the canonical character [1911.12964]. The mod \(2\) theory for real quadratic fields becomes especially explicit. For \[ K=\mathbb{Q}(\sqrt{p_1p_2\cdots p_r}), \qquad p_i\equiv 1 \pmod 4, \] let \[ T^+ := \Bigl\{(x_1,\ldots,x_r)\in (\mathbb{Z}/2\mathbb{Z})^r \ \Bigm|\ \sum_{i=1}^r x_i = 0 \Bigr\}, \] and let \(e_{ij}^+\in T^+\) be the vector with \(1\)’s in the \(i\)-th and \(j\)-th positions. Then Hirano shows \[ \operatorname{CS}_c(\rho) = \sum_{i