Crossed Product Construction

Definition

Given a von Neumann algebra $\mathcal{M}$ with modular automorphism group ${\sigma }_s$ (from Tomita-Takesaki theory), the modular crossed product is the von Neumann algebra

$\hat{\mathcal{M}}=\mathcal{M}{⋊}_{\sigma }\mathbb{R}$

constructed by adjoining to $\mathcal{M}$ a one-parameter unitary group $\{U(t){\}}_{t\in \mathbb{R}}$ implementing the modular flow: $U(t)aU(t{)}^{\ast }={\sigma }_t(a)$ for all $a\in \mathcal{M}$, $t\in \mathbb{R}$.

The crossed product $\hat{\mathcal{M}}$ is generated by:

• The original algebra $\mathcal{M}$, acting on a larger Hilbert space $\mathcal{H}\otimes L^2(\mathbb{R})$
• The unitary group $\{U(t)\}$ that translates in the “modular time” direction

Key result: If $\mathcal{M}$ is a type III${}_1$ factor, then $\hat{\mathcal{M}}$ is a type II${}_{\infty }$ factor. This is a profound algebraic upgrade:

• Type III${}_1$: no trace, no density matrix, no von Neumann entropy
• Type II${}_{\infty }$: admits a semifinite trace $\widehat{\text{Tr}}$, so von Neumann entropy $S=-\widehat{\text{Tr}}(\hat{\rho }\log \hat{\rho })$ becomes well-defined

The physical interpretation is that the crossed product “unfolds” the modular flow direction, adjoining an auxiliary degree of freedom (a “clock” conjugate to the modular Hamiltonian) that regularizes the algebra from type III to type II.

Connes’ theorem guarantees that the crossed product of a type III${}_{\lambda }$ factor is:

• Type II${}_{\infty }$ for $\lambda \in (0,1]$ (including type III${}_1$)
• Type II${}_{\infty }\otimes$ type I${}_{\infty }$ for $\lambda =0$

The construction is canonical (up to conjugacy) and does not depend on the choice of reference state $\Omega$.

Role in Research

The crossed product construction is relevant to Marcelo’s program in two distinct ways:

1. Quantifying embezzlement cost. In type III₁ algebras every normal state is an embezzling state, and the error can be made arbitrarily small without any standard entanglement monotone registering a cost. Passing to the crossed product $\hat{\mathcal{M}}$ introduces a trace, and the protocol, lifted to $\hat{\mathcal{M}}$, acquires an entropy cost that depends on the accuracy demanded — a semifinite algebra hosts no embezzling state, so the dressing is what makes the price visible. This provides a potential resolution to the question embezzlement-cost-relative-entropy: the crossed product offers a well-defined arena in which the cost of embezzlement can be computed as a von Neumann entropy difference. This is directly relevant to the PhD work under way in the group.

2. Gravitational entropy. In the holographic context reviewed by Liu (2025-liu-lectures-entanglement-vna), the crossed product of the type III${}_1$ boundary algebra with its modular automorphism group produces a type II${}_{\infty }$ algebra whose trace gives rise to the generalized gravitational entropy. This connects the Araki-Uhlmann relative entropy computed on the type III${}_1$ algebra (which is what Marcelo’s group computes for coherent, squeezed, and cat states) to a well-defined von Neumann entropy on the dressed algebra. The relationship between these two quantities is:

$S_{\text{grav}}=S_{\text{vN}}(\hat{\rho })=S_{\text{gen}}(\text{area term}+\text{bulk entanglement})$

where the area term arises from the additional “clock” degree of freedom in the crossed product.

Relations

• type-iii-von-neumann-algebras --- the crossed product is the canonical construction that promotes type III${}_1$ to type II${}_{\infty }$
• tomita-takesaki-modular-theory --- the modular automorphism group ${\sigma }_s$ is the group by which the crossed product is formed
• araki-uhlmann-relative-entropy --- relative entropy on the type III${}_1$ algebra is related to von Neumann entropy on the crossed product type II${}_{\infty }$ algebra
• entanglement-embezzlement --- the crossed product provides a framework for assigning finite cost to embezzlement protocols that are “free” at the type III${}_1$ level
• subregion-subalgebra-duality --- in holography, the crossed product of boundary subalgebras gives the gravitational algebra with well-defined entropy

Papers

• 2025-liu-lectures-entanglement-vna --- comprehensive pedagogical review of the crossed product construction and its application to gravitational entropy
• Chandrasekaran, Penington, Witten (2022-2023) --- original application of crossed products to gravitational algebras
• Witten, “Gravity and the crossed product” (2022) --- the foundational paper on gravitational crossed products

See also bell-inequalities-qft and relative-entropy-qft for the broader research context.

Notes

• The crossed product construction is not unique to type III algebras, but it has its most dramatic effect there: it promotes an algebra with no trace at all to one with a semifinite trace. For type I or type II algebras, the crossed product changes the type but does not introduce qualitatively new structure.
• Common confusion: the crossed product does not “regularize” the original algebra $\mathcal{M}$ --- it embeds it into a larger algebra $\hat{\mathcal{M}}$. The original type III${}_1$ algebra is a subalgebra of $\hat{\mathcal{M}}$, and the trace on $\hat{\mathcal{M}}$ restricts to the (trivial, divergent) weight on $\mathcal{M}$.
• The “clock” interpretation: the additional $L^2(\mathbb{R})$ factor in the crossed product Hilbert space can be thought of as a modular clock --- a reference system that breaks the modular automorphism symmetry and allows entropy to be defined. In the gravitational context, this clock is identified with the gravitational degree of freedom that makes the generalized entropy well-defined.
• Connes’ theorem on the structure of type III factors via crossed products is one of the deepest results in operator algebra theory, establishing a complete reduction of type III to type II plus an automorphism group.