Generalized Crested Products

• Generalized crested products are unified constructions that extend partial crossed products in noncommutative ring theory and Markov chain models.
• They establish exact homological sequences and spectral decompositions, linking partial cohomology with Brauer groups and representation theory.
• The framework facilitates classifying partial actions, computing eigenstructures, and integrating algebraic methods with stochastic dynamics.

Generalized crested products subsume a spectrum of algebraic and probabilistic constructions, unifying partial generalized crossed products in noncommutative ring theory with the generalized crested products of Markov kernels on product spaces. These frameworks extend classical crossed product and wreath product constructions, providing exact homological sequences, spectral decompositions, and deep connections to representation and cohomology theories (Dokuchaev et al., 2021, D'Angeli et al., 2010).

1. Partial Generalized Crossed Products: Algebraic Formalism

Let $R$ denote a unital (not necessarily commutative) ring with center $Z$. A partially invertible $R$-bimodule $P$ is finitely generated and projective both as a left and right $R$-module, with evaluation maps $R \to \operatorname{End}_R(P)$ (given by left and right multiplication) surjective. The Picard semigroup $\mathbf{PicS}(R)$ comprises isomorphism classes $[P]$ of such bimodules, with multiplication $[P] \cdot [Q] := [P \otimes_R Q]$ and unit $[R]$. The group of units coincides with the Picard group of invertible bimodules.

A unital partial representation $\Theta: G \to \mathbf{PicS}(R)$ of a group $G$ satisfies:

• $[\Theta_1]=[R]$
• $[\Theta_x][\Theta_{x^{-1}}][\Theta_x]=[\Theta_x]$
• $$[\Theta_x][\Theta_y][\Theta_{y^{-1}}]=[\Theta_{xy}][\Theta_{y^{-1}}]$$
• $$[\Theta_{x^{-1}}][\Theta_x][\Theta_y]=[\Theta_{x^{-1}}][\Theta_{xy}]$$ Unitality additionally requires $[\Theta_x][\Theta_{x^{-1}}] = [R 1_x]$ for some central idempotent $1_x \in Z$.

A partial generalized crossed product (termed generalized crested product by Dokuchaev–Rocha) is constructed as follows. For each $x, y \in G$, an $R$-bimodule isomorphism (factor map) $$f_{x,y} : \Theta_x \otimes_R \Theta_y \to R 1_x\,\Theta_{xy}$$ is given, subject to associativity constraints. The direct sum $A = \bigoplus_{x \in G} \Theta_x$ becomes an associative unital $R$-algebra with multiplication $u_x \star u_y := f_{x,y}(u_x \otimes u_y)$. Equivalence classes of such products, under compatible isomorphisms, form an abelian group $\mathcal{C}(\Theta / R)$ with multiplication $$[A] \cdot [A'] = [\bigoplus_x \Theta_x \otimes_R \Theta'_x]$$ (Dokuchaev et al., 2021).

2. Partial Cohomology and the Brauer Group Structure

The subgroup $\mathcal{C}_0(\Theta / R)$ of $\mathcal{C}(\Theta / R)$, consisting of products with underlying bimodules isomorphic to the original $\Theta_x$, is isomorphic to the second partial cohomology group $H^2_p(G, \Theta / R)$ where the $G$-action on $Z$ is induced by the partial representation. This identification corresponds to the triviality of 2-cocycle obstructions up to coboundary. The quotient $$\mathcal{B}(\Theta / R) = \mathcal{C}(\Theta / R)/\operatorname{Im}[\operatorname{Pic}_Z(R)^{(G)} \to \mathcal{C}(\Theta / R)]$$ is the Brauer group of equivalence classes of partial generalized crossed products (Dokuchaev et al., 2021).

3. The Seven-Term Exact Sequence

Given a unital ring extension $R \subseteq S$ with equal unities and a unital partial representation $\Theta: G \to \mathcal{S}_R(S)$ (the monoid of $R$-subbimodules of $S$), the construction yields a canonical seven-term exact sequence:

$$1 \to H^1_p(G, Z) \to P_Z(A/R)^{(G)} \to \operatorname{Pic}_Z(R) \cap \operatorname{PicS}_Z(R)^{\Theta^*} \to H^2_p(G, Z) \to \mathcal{B}(A/R) \to \overline{H}^1_p(G, \operatorname{PicS}_0(R)) \to H^3_p(G, Z) \to 1$$

Each term and map in the sequence has an explicit algebraic and cohomological interpretation:

• $H^n_p(G, Z)$: Partial cohomology groups;
• $P_Z(A/R)^{(G)}$: G-invariant relative Picard classes split by $A$;
• $$\operatorname{Pic}_Z(R) \cap \operatorname{PicS}_Z(R)^{\Theta^*}$$: Z-central invertible bimodules fixed under $\Theta$'s partial action;
• $\mathcal{B}(A/R)$: Brauer group as above;
• $\overline{H}^1_p(G, \operatorname{PicS}_0(R))$: Quotient of first partial cohomology with semigroup coefficients (Dokuchaev et al., 2021).

Table of Key Groups and Their Roles:

Symbol	Description
$\mathbf{PicS}(R)$	Picard semigroup of partially invertible bimodules
$H^2_p(G, \Theta/R)$	Second partial cohomology, classifying crossed products
$\mathcal{B}(A/R)$	Brauer group of partial generalized crossed products

4. Generalized Crested Product of Markov Chains

A distinct but related structure is the generalized crested product of Markov chains. Given finite state spaces $X_i$ with irreducible, reversible Markov kernels $P_i$ and a finite poset $(I, \leq)$, the generalized crested product is the Markov operator $\mathcal{P}$ on $X = \prod_{i \in I} X_i$ defined by: $$\mathcal{P}((x_1, ..., x_n), (y_1, ..., y_n)) = \sum_{i \in I} p^0_i \; p_i(x_i, y_i) \left(\prod_{j \in H(i)} U_j(x_j, y_j)\right) \left(\prod_{j \notin H[i]} I_j(x_j, y_j)\right)$$ where $H(i) = \{j \in I: j < i\}$, $U_j$ is the uniform kernel on $X_j$, and $I_j$ is the identity kernel (D'Angeli et al., 2010). At each step, coordinate $i$ is updated via $P_i$, coordinates $j < i$ are resampled uniformly, and all other coordinates are fixed.

5. Spectral Theory and Representation Theoretic Interpretation

For the generalized crested product of Markov chains, a complete spectral decomposition is available. If each $P_i$ is reversible,

$$L(X_i) = \bigoplus_{j_i = 0}^{r_i} V^i_{j_i}, \quad P_i|_{V^i_{j_i}} = \lambda_{j_i} \mathrm{Id}$$

then $L(X)$ decomposes as

$$L(X) = \bigoplus_{S \in \mathcal{S}} \bigoplus_{\underline{j}} W_{S, \underline{j}}$$

where $S$ runs over antichains of $(I, \leq)$, and $W_{S, \underline{j}}$ are explicit eigenspaces with corresponding eigenvalues

$$\lambda_{S, \underline{j}} = \sum_{h=1}^k p_{i_h}^0 \lambda_{j_{i_h}} + \sum_{i \notin A[S]} p_i^0$$

with $A(S)$ the poset successors of all elements in $S$ (D'Angeli et al., 2010).

The spectral decomposition mirrors the decomposition into irreducible modules for the action of the generalized wreath product of permutation groups $G = \prod_{i \in I} \operatorname{Sym}(X_i)$ structured by the poset. If $K \subset G$ stabilizes a base point, $(G, K)$ forms a Gelfand pair and the eigenspaces correspond to irreducible $G$-modules, with spherical functions as normalized eigenfunctions of the chain.

6. Special Cases, Examples, and Connections

Crossed Products and Classical Results

• For global actions $\alpha: G \to \operatorname{Aut}(R)$ and $2$-cocycles $U: G \times G \to U(R)$, one obtains the classical (twisted) crossed product $R^*_{\alpha,U} G$ as a special case (Dokuchaev et al., 2021).
• For commutative $R$ and a partial action by idempotent-preserving automorphisms, the construction recovers partial Galois theory.

Markov Chain Models

• The Ehrenfest diffusion model appears as the case of a total chain poset, binary state spaces, and flip kernels.
• The insect Markov chain on an ultrametric tree emerges when all $P_i$ are uniform kernels, providing transitions governed by up–down tree dynamics.

Concrete Examples

For $I = \{1,2\}$ with no order: $$X = X_1 \times X_2, \quad \mathcal{P} = p_1^0\,(P_1 \otimes I_2) + p_2^0\,(I_1 \otimes P_2)$$ For $I = \{1<2\}$: $$\mathcal{P} = p_1^0\,(P_1 \otimes I_2) + p_2^0\,(P_2 \otimes U_1)$$ Characteristic spectral distributions are computable, matching the antichain decomposition (D'Angeli et al., 2010).

7. Significance and Context

Generalized crested products encapsulate a range of phenomena at the intersection of ring theory, group cohomology, Markov processes, and representation theory. The algebraic framework provides a systematic avenue to classify and understand partial actions, twisted algebras, and cohomological obstructions in noncommutative settings, including exact sequences extending classical Galois and Brauer theory (Dokuchaev et al., 2021). The probabilistic formulation unifies and generalizes well-known Markov and diffusion models, with complete spectral and combinatorial characterizations, and situates them in the context of group representations and symmetric structures. These dual perspectives enable a rigorous analysis of partial symmetries, module structures, and stochastic dynamics in both algebraic and probabilistic domains (D'Angeli et al., 2010).