Sem II Week 10 — Liu Lectures II: Crossed Products, Semiclassical Limits, Algebraic ER=EPR

Week 9 installed two pillars: type III at large $N$, and modular flow as a candidate bulk geometric flow. This week completes the Liu block by synthesizing the continuous core, gravitational dressing, and Liu’s conditional algebraic ER=EPR proposal. We will be deliberately precise about the arrows. The crossed product is an exact algebraic construction; the interpretation of its auxiliary variable as an ADM clock is model-dependent; and Liu’s connectivity proposal is an “if and only if” classification only inside a stated semiclassical, non-stringy domain. The familiar standard-pair identity $J{\mathcal{A}}_{R}J={\mathcal{A}}_R^{\prime }$ is not that proposal.

0. Reading

Primary:

• Hong Liu, arXiv:2510.07017, §V (crossed product by modular flow), §VII (subregion–subalgebra duality and islands), §VIII, especially §VIII.3 (emergent connectivity and algebraic ER=EPR), and §IX (algebraic approaches to quantum-gravity regimes).

Secondary / gentler:

• Sem I Wks 13–14 (the abstract crossed product and dressed entropy).
• Sem II Wks 2–3, 6–7 (Witten and CPW crossed-product constructions).

Optional research reading:

• Chandrasekaran, Penington, Witten, arXiv:2209.10454 (the eternal-BH crossed product Liu §§V and IX place in a wider framework).
• Witten, “Gravity and the crossed product,” arXiv:2112.12828 §§3.2–3.5 (with §4 for the extension to other conserved charges).
• Maldacena, Susskind, “Cool horizons for entangled black holes,” arXiv:1306.0533 (the original ER=EPR proposal).

0.1 How to use this master dossier

• Classroom core: §§1–2, 3.1, and 3.3, with Problems 1–4 and 7. This route derives the core, separates the Witten and CPW limits, and states Liu’s actual three-branch proposal with its domain.
• Full derivation / self-study: §§3.2, 3.4, and 4–6, with Problems 5, 6, and 8. This route checks the finite-dimensional standard pair, trace normalization, research map, and AAJ handoff.
• Research extension: Problems 9–10. The deliverable is a source-scope table or long-form exposition that distinguishes exact algebra, limiting assumptions, and geometric interpretation.

For this week, remembering a slogan is not enough. The student must be able to explain why a standard-pair commutant identity and Liu’s algebraic ER=EPR proposal answer different questions.

1. The crossed product as “adding a clock”

1.1 Liu’s framing

We built the modular crossed product abstractly in Sem I Wk 13 and applied it in Blocks 1–2. Liu §V reframes it in a way worth adopting because it makes the physics of the construction transparent.

Start from a type III${}_1$ boundary algebra $\mathcal{A}$ with modular flow ${\sigma }_t^{\omega }$. Its modular action is nontrivial in the outer automorphism group; in particular, there is no unitary group inside $\mathcal{A}$ implementing the full flow. In the large-$N$ gravitational examples, the missing implementer is related to an energy collective coordinate.

The crossed product

$$\hat{\mathcal{A}}=\mathcal{A}{⋊}_{{\sigma }^{\omega }}\mathbb{R}$$

adjoins exactly such a clock: an auxiliary degree of freedom on $L^2(\mathbb{R})$, with the translation $\lambda (t)$ implementing the modular flow as an inner automorphism of $\hat{\mathcal{A}}$. Liu’s slogan: the crossed product makes the modular flow inner by adjoining the clock that reads modular time.

The slogan is worth checking rather than repeating, and it takes three lines. Recall the concrete realization from Sem I Wk 13 §2.1: $\hat{\mathcal{A}}$ acts on $\mathcal{H}\otimes L^2(\mathbb{R})$, and is generated by

$$(\pi (a)\xi )(q)={\sigma }_{-q}(a)\xi (q),(\lambda (t)\xi )(q)=\xi (q-t),$$

for $a\in \mathcal{A}$ and $t\in \mathbb{R}$, where $\xi$ is an $\mathcal{H}$-valued function of the regular coordinate $q$. Here $(Q\xi )(q)=q\xi (q)$, $P=-i{\partial }_q$, and $\lambda (t)=e^{-itP}$. The Fourier spectral variable of $P$ is $p$; the parameter of the dual action will be $r$; and a physical time or rapidity $u$ is related to modular time only after a model supplies a scale, such as $u={\beta }_{H}t$. Note that $\pi$ is a faithful representation of $\mathcal{A}$ and $\lambda$ a unitary representation of the translations. Conjugating one by the other, and using $(\lambda (t{)}^{\ast }\xi )(q)=\xi (q+t)$,

$$(\lambda (t)\pi (a)\lambda (t{)}^{\ast }\xi )(q)=(\pi (a)\lambda (t{)}^{\ast }\xi )(q-t)={\sigma }_{-(q-t)}(a)\xi (q)={\sigma }_{t-q}(a)\xi (q),$$

while acting directly with the flowed element gives $(\pi ({\sigma }_t(a))\xi )(q)={\sigma }_{-q}({\sigma }_t(a))\xi (q)={\sigma }_{t-q}(a)\xi (q)$. The two agree, so

$$\boxed{\lambda (t)\pi (a)\lambda (t{)}^{\ast }=\pi ({\sigma }_t(a))\text{for all }a\in \mathcal{A},t\in \mathbb{R}.}$$

The modular flow is implemented by conjugation with an element of $\hat{\mathcal{A}}$ — it has become inner. [Proved.]

Were ${\sigma }_t=\mathrm{Ad}(v_t)$ for a unitary representation $v_t$ inside $\mathcal{A}$, the action would already be inner and the crossed product would reduce to the inner-action tensor-product model. The III${}_1$ case is different: the continuous core is a II${}_{\infty }$ factor.

Physical picture. The crossed product supplies an internal implementer for a flow that was outer on the original algebra. Calling the auxiliary canonical pair a clock is natural when a physical construction identifies one of its variables with energy and the conjugate one with time. Witten and CPW provide such an identification in particular large-$N$ black-hole limits. Outside those settings, “clock” remains an interpretation of the group representation, not a theorem about gravity.

1.2 Gauging gravity

In relational models the crossed product can be realized as a constraint-invariant algebra of system plus clock. In asymptotically AdS gravity, however, the ADM Hamiltonian is a boundary charge rather than a constraint that vanishes on all states. The precise Witten/CPW statement is that including the energy collective coordinate and its conjugate timeshift realizes the continuous core.

The one-line summary, which we adopt as the block’s thesis:

In the Witten/CPW regime, the relevant semiclassical exterior algebra is realized as the continuous core of the strict-$N$ type III${}_1$ algebra.

2. The semiclassical reduction

2.1 The statement

Liu §§V and IX organize the entropy statement that Blocks 1–2 used. There is, however, no single undifferentiated “large-$N$ limit” behind both constructions.

Semiclassical reduction. [Stated only — hypothesis-explicit; refs: Liu §§V, IX; CPW §§3–4, with Witten §§3.4–3.5 for the trace framework.] For the controlled class of gravitational states to which the CPW calculation applies, the entropy of the corresponding normal state on the dressed algebra agrees, to the stated semiclassical order, with generalized entropy:

$$S_{\mathrm{vN}}(\hat{\rho })⟶\frac{A_{\mathrm{RT}}}{4G_N}+S_{\mathrm{out}}+\mathrm{const},$$

where $A_{\mathrm{RT}}$ is the area of the RT/horizon surface bounding the entanglement wedge, $S_{\mathrm{out}}$ is the bulk matter entropy outside it, and the state-independent constant reflects the trace-rescaling ambiguity (Sem I Wk 14 §1.3).

The structural derivation combines the exact semifinite entropy identity of Sem I Wk 14 with a further gravitational input: the modular-energy and relative-entropy terms must be identified with the area response and bulk entropy. Witten supplies the crossed-product trace framework and a black-hole entropy normalization; CPW performs the explicit generalized-entropy matching in its controlled microcanonical regime. Liu presents these results in a common language; the crossed product by itself does not prove the geometric identification for an arbitrary spacetime.

Physical picture. The generalized entropy combines matter correlations with a gravitational area term, but neither term is simply “the entropy of one tensor factor.” In CPW the ADM-energy distribution, bulk relative entropy, and area response appear as separate contributions to one type-II density. Their recombination into $A/(4G_N)+S_{\mathrm{out}}$ uses the semiclassical horizon equations.

2.2 Auditing the reduction in three layers

The safest calculation keeps the algebraic model, the QFT regulator, and the gravitational identification separate.

Layer 1 — exact Fourier model. In the inner-action laboratory of Sem I Week 14, a system state $\rho$ and a probability distribution $\mu (p)$ in the spectral variable of $P$ have

$$S_{\hat{\tau }}=S(\rho )+H(\mu )-{\mathbb{E}}_{\mu }[p].$$

For a Gaussian of mean $p_0$ and width $\sigma$,

$$\boxed{S_{\hat{\tau }}=S(\rho )+\frac{1}{2}\log (2\pi e{\sigma }^2)-p_0.}$$

This is an exact trace calculation, but the inner-action crossed product is not the type III${}_1$ Rindler core.

Layer 2 — regulated matter QFT. The matter entropy is computed independently. In $1+1$ dimensions it is logarithmic; in $3+1$ dimensions its leading term is $c_{2}A/{ϵ}_{\mathrm{UV}}^2$. The cutoff ${ϵ}_{\mathrm{UV}}$ is not $\sigma$.

Layer 3 — semiclassical gravity. Witten identifies the energy collective coordinate and interprets the entropy normalization relative to a black hole. CPW then uses the first law plus horizon relative entropy to obtain

$$S_{\mathrm{alg}}=\frac{A}{4G_N}+S_{\mathrm{out}}+\mathrm{const}$$

in their stated regimes. The operator algebra alone does not determine $G_N$ or the area coefficient.

Notice the sign in Layer 1: as $\sigma \to 0$, the differential entropy tends to $-\infty$. The positive matter UV divergence has a different origin and, in gravity, is absorbed together with the renormalization of the area coupling.

Physical picture. The crossed product supplies a trace and therefore a controlled notion of renormalized entropy. Gravity tells us what the collective coordinate means and how its mean energy changes the horizon area. Matter QFT supplies the short-distance divergence. These are cooperating ingredients, not three names for the same regulator.

2.3 Witten and CPW take different routes to the core

This distinction is easy to lose because the final operator algebra is the same continuous core.

Witten’s canonical-ensemble route. At leading strict large $N$, the rescaled energy fluctuation $U=(H_R-⟨H_R⟩)/N$ becomes central relative to the single-trace algebra. If one stopped there, the algebra would have a classical $L^{\infty }$ energy sector. Witten retains the perturbative $1/N$ action that couples this collective coordinate to the modular generator. It is this retained correction—not the leading central algebra by itself—that realizes the type II${}_{\infty }$ crossed product. The construction is therefore a perturbative large-$N$ algebra; it is not a claim that the exact finite-$N$ CFT algebra has become type II.

CPW’s microcanonical route. CPW first choose a microcanonical TFD window centered at $E_0=O(N^2)$ and keep the subtracted energy $h_R=H_R-E_0$ with $O(1)$ fluctuations as $N\to \infty$. In this scaling $h_R$ acts nontrivially on the limiting type III${}_1$ algebra, so the right algebra is already the type II${}_{\infty }$ core in the strict microcanonical limit. Their “semiclassical state” condition then concerns a separate narrow distribution in the conjugate timeshift, controlled by a parameter $\epsilon$ that is taken small only after the algebraic limit has been defined.

These are not competing answers. Witten explains how the noncentral collective coordinate appears when the leading canonical large-$N$ algebra is corrected; CPW choose a microcanonical scaling in which the appropriate $O(1)$ generator survives from the outset and can be used in an explicit entropy calculation. The common continuous-core structure is exact once the corresponding action has been specified. The gravitational interpretation and the error estimates retain the hypotheses of the route used to obtain it.

3. Algebraic ER=EPR: exact preliminaries and Liu’s conditional proposal

3.1 The standard-form pattern

Before stating Liu §VIII.3, recall an exact modular fact that is often confused with the proposal. A standard pair $({\mathcal{A}}_R,\Omega )$ gives a canonical antiunitary $J$ and a commutant. If the physical left algebra has separately been identified as that commutant, then the two-sided modular pattern follows.

Two-sided modular pattern. [Exact algebraic statement.] Let $\Omega$ be cyclic and separating for ${\mathcal{A}}_R$, and suppose the physical left algebra is represented as ${\mathcal{A}}_L={\mathcal{A}}_R^{\prime }$. Tomita–Takesaki theory then gives

$$J{\mathcal{A}}_{R}J={\mathcal{A}}_L={\mathcal{A}}_R^{\prime }.$$

The vector is cyclic and separating for the one-sided algebra ${\mathcal{A}}_R$, not for the joined algebra ${\mathcal{A}}_R\vee {\mathcal{A}}_L$. In an irreducible two-sided representation the joined algebra may be $\mathcal{B}(\mathcal{H})$, for which no vector is separating when $\dim \mathcal{H}>1$.

The equality $J{\mathcal{A}}_{R}J={\mathcal{A}}_R^{\prime }$ is automatic once $\Omega$ is cyclic and separating; by itself it cannot diagnose a wormhole. The holographic content lies in identifying that commutant with the opposite boundary algebra inside a code subspace and in establishing what sort of bulk dual the state has. In particular, this standard-pair identity is not the algebraic ER=EPR classification below.

3.2 Model proof: the finite-dimensional TFD

The structure is visible in finite dimensions, where every object is a matrix. [Model proof — finite dimensions.]

Take $\mathcal{H}={\mathcal{H}}_R\otimes {\mathcal{H}}_L$ with ${\mathcal{H}}_R\cong {\mathcal{H}}_L\cong {\mathbb{C}}^n$, and

$${\mathcal{A}}_R=\mathcal{B}({\mathcal{H}}_R)\otimes 1,{\mathcal{A}}_L=1\otimes \mathcal{B}({\mathcal{H}}_L).$$

The TFD vector at inverse temperature $\beta$ for a Hamiltonian with eigenvalues $E_j$ is

$$|\Psi ⟩=\sum _j\sqrt{p_j}|j{⟩}_R\otimes |j{⟩}_L,p_j=\frac{e^{-\beta E_j}}{Z},$$

which is cyclic and separating for ${\mathcal{A}}_R$ precisely when every $p_j>0$ — the finite-dimensional shadow of “the TFD is a faithful state.”

The modular conjugation. From Sem I Wk 15 / Sem II Wk 5, $\Delta =\rho \otimes {\rho }^{-1}$ with $\rho =\mathrm{diag}(p_j)$, and $J$ acts on the computational basis by swapping the two factors and conjugating coefficients:

$$J(\sum _{k,l}{c}_{kl}|k{⟩}_R\otimes |l{⟩}_L)=\sum _{k,l}\overline{c_{kl}}|l{⟩}_R\otimes |k{⟩}_L.$$

The mirror property, computed. Let $a\in \mathcal{B}({\mathcal{H}}_R)$ have matrix elements $a_{ml}$. Applying $J$, then $a\otimes 1$, then $J$ again to a general vector $\sum c_{kl}|kl⟩$:

$$J(a\otimes 1)J\sum _{k,l}{c}_{kl}|kl⟩=J\sum _{k,l,m}\overline{c_{kl}}{a}_{ml}|mk⟩=\sum _{k,l,m}{c}_{kl}\overline{a_{ml}}|km⟩,$$

while acting directly with $1\otimes \bar{a}$, where $\bar{a}$ is the entrywise complex conjugate, gives

$$(1\otimes \bar{a})\sum _{k,l}{c}_{kl}|kl⟩=\sum _{k,l,m}{c}_{kl}\overline{a_{ml}}|km⟩.$$

The two agree for every $c_{kl}$, so

$$J(a\otimes 1)J=1\otimes \bar{a}\text{for all }a\in \mathcal{B}({\mathcal{H}}_R).$$

Since $a\mapsto \bar{a}$ is a bijection of $\mathcal{B}({\mathcal{H}}_R)$ onto $\mathcal{B}({\mathcal{H}}_L)$, this gives $J{\mathcal{A}}_{R}J={\mathcal{A}}_L$. And ${\mathcal{A}}_R^{\prime }=(\mathcal{B}({\mathcal{H}}_R)\otimes 1{)}^{\prime }=1\otimes \mathcal{B}({\mathcal{H}}_L)={\mathcal{A}}_L$ by the commutation theorem for tensor products. Both halves of the criterion hold:

$$J{\mathcal{A}}_{R}J={\mathcal{A}}_L={\mathcal{A}}_R^{\prime }.\square$$

And how standardness fails. Now take a factorized pure state $|{\psi }_R⟩\otimes |{\psi }_L⟩$. Then

$${\mathcal{A}}_R(|{\psi }_R⟩\otimes |{\psi }_L⟩)={\mathcal{H}}_R\otimes |{\psi }_L⟩,$$

a proper subspace of $\mathcal{H}$ whenever $\dim {\mathcal{H}}_L>1$. The vector is neither cyclic nor separating: any nonzero $a$ annihilating $|{\psi }_R⟩$ gives $(a\otimes 1)|{\psi }_R{\psi }_L⟩=0$. The standard-form construction fails at its first hypothesis.

A mixed product density ${\rho }_R\otimes {\rho }_L$ defines a perfectly good normal state, but it is not a vector state on the original bipartite Hilbert space. A purification introduces ancillary factors. In that enlarged representation, the commutant of the represented right algebra contains both the original left operators and the relevant ancilla operators, so the original boundary algebra ${\mathcal{A}}_L$ is generally smaller than the full commutant. This is a reminder that the commutant statement concerns a specified representation; it is not a state-independent label for geometry.

Physical picture. The finite-dimensional check tests faithfulness of the one-sided state: every Schmidt coefficient must be nonzero. It does not test maximal entanglement, since unequal positive Schmidt coefficients still give a cyclic-separating vector. Whether those modular data admit a smooth connected bulk interpretation is a further holographic question.

3.3 Liu §VIII.3: the proposal and its domain

Liu reviews the algebraic ER=EPR proposal of Engelhardt and Liu. It is a conditional proposal, not a theorem of Tomita–Takesaki theory. Its starting assumptions already contain substantial bulk information:

• two entangled systems $R_1$ and $R_2$ are in a pure semiclassical state;
• the state has a gravitational bulk dual ${\mathcal{W}}_{R_1{R}_2}$ in the $G_N\to 0$ limit;
• every part of that bulk is connected to at least one asymptotic boundary, excluding baby-universe components;
• the discussion is in the ${\alpha }^{\prime }\to 0$ regime, where the geometric notions used in the classification are meaningful.

Let ${\mathcal{M}}_{R_1}$ and ${\mathcal{M}}_{R_2}$ be the corresponding large-$N$ boundary algebras. Within this domain, the proposal is:

1. ${\mathcal{W}}_{R_1{R}_2}$ is disconnected if and only if both boundary algebras are type I.
2. It has a classical wormhole connecting the boundaries if and only if both algebras are type III${}_1$ and ${\mathcal{W}}_{R_1{R}_2}$ is classical.
3. It has a quantum wormhole if and only if ${\mathcal{W}}_{R_1{R}_2}$ is quantum volatile and neither boundary algebra is type I.

The word “classical” in item 2 is not decorative. Liu calls the geometry classical when, as $G_N\to 0$, fluctuations of diffeomorphism-invariant observables vanish as $O(G_N^a)$ for some $a>0$, while invariant lengths, areas, and volumes do not themselves grow as $O(G_N^{-a})$. A geometry is quantum volatile when one or both conditions fail. An evaporating black-hole interior whose proper length grows like $O(G_N^{-1})$ is the paradigm: even when local fluctuations appear controlled, a parametrically large invariant makes the limit geometrically volatile.

This definition repairs a weakness of the slogan “enough entanglement implies a bridge.” The amount of entanglement alone does not distinguish a smooth classical bridge from a long, fluctuating quantum region, nor does it exclude disconnected entangled phases. The algebra type records the structure of entanglement, while the classical/volatile condition records the behavior of the bulk limit.

Two examples make the three branches concrete.

• For the low-temperature TFD below the Hawking–Page transition, the large-$N$ boundary algebras are type I in Liu’s account. The two bulks are disconnected despite nonzero entanglement; the proposal assigns neither a classical nor a quantum wormhole.
• In the evaporating-black-hole example, the boundary algebra may remain type III${}_1$ on both sides of the Page transition. Before the Page time the relevant entanglement wedge is quantum volatile and the proposal calls the connection a quantum wormhole. After the Page time the wedge becomes classical and the connection is through a classical wormhole anchored at the quantum extremal surface. Algebra type alone does not distinguish these two stages; the volatility condition is essential.

Finally, Liu explicitly restricts this classification to ${\alpha }^{\prime }\to 0$. In a stringy regime, the geometric concepts entering “classical wormhole” and “quantum volatile” need revision, and type III${}_1$ alone is not enough to infer connectivity. Thus even the proposal’s if-and-only-if language must be carried together with its domain.

3.4 What the standard-pair calculation does—and does not—test

The finite-dimensional calculation in §3.2 tests whether a one-sided density is faithful: full Schmidt rank makes the vector cyclic and separating. It does not determine the large-$N$ algebra type, the $G_N\to 0$ scaling of bulk invariants, or the ${\alpha }^{\prime }$ regime. Conversely, Liu’s three-way proposal does not replace the commutant construction needed to describe two exterior observable algebras.

The two ideas should therefore be used at different stages:

$$\text{standard pair}⟹\text{exact modular mirror/commutant structure},$$

whereas

$$\text{large-N algebra types + bulk classicality/volatility + domain assumptions}⟹\text{the proposed connectivity classification}.$$

One last warning is orthogonal to both statements. CHSH violation is a state-and-observable property. Type III${}_1$ structure permits strong violations under additional hypotheses, but neither algebra type nor cyclic-separating property forces a specified quartet to saturate $2\sqrt{2}$.

4. A course research map prompted by Liu

The single most useful page of the block: where the three research papers sit.

                    type III_1 boundary algebra  A  (strict large-N limit)
                                   |
                                   |  crossed product by modular flow
                                   |  (ADM-clock meaning requires the Witten/CPW input)
                                   v
                    type II_infty continuous core  Â
                                   |
                                   |  gravitational matching in a controlled regime
                                   |  S_alg  =  A/4G_N + S_out + const
                                   v
        ┌──────────────────────────┼───────────────────────────┐
        |                          |                            |
   one-sided                   two-sided                    unitary deformation
   (Witten 2022)               (CPW 2022)                   (AAJ 2025)
   retained 1/N action         strict microcanonical N→∞    GJW input
   right exterior core         right core + commutant       weight/Jacobian + BCH
   BH entropy normalization    controlled S_gen matching    type-II entropy series
        |                          |                            |
        └────────── related uses of the crossed-product framework ────────────┘
                                   |
                            de Sitter sibling (CLPW, Block 5 Wk 14):
                            invariant observer algebra with a positive
                            observer Hamiltonian  ⇒  type II_1 (finite trace)

The reading:

1. Witten 2022 (Block 1): the right exterior algebra with its energy collective coordinate, after retaining the perturbative action that is invisible in the leading central large-$N$ algebra. Output: a type II${}_{\infty }$ factor, trace entropy up to a constant, and its normalization relative to the reference black-hole entropy.
2. CPW 2022 (Block 2): the right core together with its dressed commutant in the strict microcanonical large-$N$ setting. Output: the explicit $S_{\mathrm{gen}}$ matching for CPW’s semiclassical states and the two-sided modular pattern.
3. AAJ 2025 (Block 4): unitary transport of the algebra–state system, with a GJW deformation as the application. AAJ expand the deformation-dependent crossed-product weight and spectral Jacobian using BCH nested commutators. Through $O(1/N^2)$ their source-controlled ledger contains the GJW linear structure, five additional linear structures, and fifteen quadratic structures. This is a type-II von Neumann entropy expansion; it is not automatically a calculation of a bulk time advance or a term-by-term proof of $\delta S_{\mathrm{alg}}=\delta S_{\mathrm{gen}}$.
4. CLPW (Block 5 Wk 14 aside): a closely related observer construction in de Sitter gives a type II${}_1$ algebra. The finite trace follows from the detailed invariant-observer setup, including the positive observer Hamiltonian; compactness of the horizon alone is not an operator-algebraic proof of the type.

Physical picture. The diagram is a map, not an equivalence theorem. The continuous core supplies the type II${}_{\infty }$ algebraic backbone in the Witten/CPW setting. The identification of its entropy with generalized entropy uses additional semiclassical dynamics. AAJ transports and reweights a deformed algebra–state system, while CLPW is a related invariant-observer construction whose trace is finite. Seeing the common pattern is useful precisely because it also makes the different hypotheses visible.

5. What the Liu block prepares us to compare

Liu §§VIII–IX go beyond an unperturbed entropy formula: they use algebra type, modular structure, subregion duality, and observer dressing to discuss classical and quantum-volatile gravitational regimes. Block 4 asks a different, more narrowly perturbative question: how does a type-II crossed-product entropy change when the full algebra–state system is transported by a specified unitary deformation?

The application is the Gao–Jafferis–Wall (GJW) double-trace deformation, with a sign and switching profile chosen so that the bulk calculation produces negative averaged null energy and a traversable window. AAJ use the corresponding interaction as an input to their unitary algebraic framework. Their entropy series comes from the changed crossed-product weight, a spectral Jacobian, and BCH nested commutators. It is not a Connes-cocycle expansion. The Araki/Connes cocycle remains a valuable course comparison for fixed-algebra state perturbations, but the two perturbative machines must not be identified.

Students enter Block 4 with:

1. the structural picture from Liu (this block);
2. explicit dressed-entropy computations from Witten and CPW (Blocks 1–2);
3. the Casini–Huerta–Myers / boundary-modular-flow dictionary (Wk 9);
4. the Connes cocycle as a comparison formalism whose domain must be kept distinct from AAJ’s unitary transport.

6. What to take away

• Crossed product and clock: the crossed product makes the outer modular flow inner. Its canonical pair becomes an ADM-energy/timeshift pair only after the physical identification made in the Witten/CPW regime. For a type III${}_1$ factor the continuous core is type II${}_{\infty }$.
• Semiclassical reduction (stated only): the exact Fourier laboratory gives $S_{\hat{\tau }}=S(\rho )+H(\mu )-{\mathbb{E}}_{\mu }[p]$ and, for a Gaussian, $S(\rho )+\frac{1}{2}\log (2\pi e{\sigma }^2)-p_0$. The further equality with $A/(4G_N)+S_{\mathrm{out}}+\mathrm{const}$ is a gravitational matching. The clock-width term is not the matter area-law divergence.
• Standard pair versus algebraic ER=EPR: a faithful TFD-like vector gives the exact relation $J{\mathcal{A}}_{R}J={\mathcal{A}}_R^{\prime }={\mathcal{A}}_L$. Liu’s §VIII.3 proposal is different: within a pure semiclassical bulk domain and the ${\alpha }^{\prime }\to 0$ limit, it classifies disconnected, classical-wormhole, and quantum-volatile-wormhole regimes using boundary algebra types together with bulk classicality.
• Research map: Witten, CPW, AAJ, and the CLPW aside use related operator-algebraic structures with different physical inputs; their conclusions should not be transferred without their hypotheses.
• Handoff: AAJ use unitary covariance, a deformation-dependent weight and Jacobian, and BCH—not the course’s Connes-cocycle series—to organize the deformed type-II entropy.

7. Looking ahead

Block 4 (Weeks 11–13) studies Ahmad–Jefferson’s unitary algebraic perturbation theory. Week 11 separates AAJ’s transported algebra–state system from an Araki/Connes fixed-algebra comparison and introduces the GJW application. Week 12 audits AAJ’s actual linear and quadratic term ledgers, then builds an exact finite-matrix model and a reproducible finite-regulator field-theory protocol. Week 13 separates AAJ’s stated outlook from course-generated research questions. The final-write-up draft from this week’s problem set will be refined throughout Block 4.

8. Problem set

Core problems.

1. The dual action, and what it remembers. Let $q$ be the regular coordinate, let $(Q\xi )(q)=q\xi (q)$, and let the dual action have parameter $r$:

$${\theta }_r(\pi (a))=\pi (a),{\theta }_r(\lambda (t))=e^{itr}\lambda (t).$$

In the regular representation it is implemented ambiently by $e^{irQ}$. Verify both formulas directly and show that the fixed-point algebra is $\pi (\mathcal{A})$. Then distinguish this fixed-point statement from Takesaki duality, which says

$$(\mathcal{A}{⋊}_{\sigma }\mathbb{R}){⋊}_{\theta }\mathbb{R}\cong \mathcal{A}\overline{\otimes }\mathcal{B}(L^2(\mathbb{R})).$$

Explain what information the dual action retains about the extension. Do not conclude that a physically identified clock is automatically gauge redundancy.

2. A different clock distribution. In the exact Fourier model take

$${\mu }_{\eta }(p)=\eta e^{-\eta p}{1}_{p\ge 0},\eta >0.$$

Check that $H({\mu }_{\eta })=1-\log \eta$ and $\mathbb{E}[p]=1/\eta$. Hence compute

$$S_{\hat{\tau }}=S(\rho )+H({\mu }_{\eta })-\mathbb{E}[p].$$

Which part follows from the trace weight $e^{-p}dp$, and which part depends on the shape of the probability distribution? Compare its standard deviation $1/\eta$ with the Gaussian width $\sigma$.

3. What the distribution parameters do. Using the boxed result of §2.2, compute $\partial S_{\hat{\tau }}/\partial \sigma$ and $\partial S_{\hat{\tau }}/\partial p_0$. Explain why the answers are $1/\sigma$ and $-1$, respectively. Then explain why neither derivative computes the matter area-law divergence, and why identifying $p_0$ with a horizon-energy variable requires an extra physical dictionary.

4. The research map from memory. Reproduce §4’s diagram and write one sentence per node (Witten / CPW / AAJ / CLPW) stating which operator-algebraic framework it uses, which extra physical input it adds, and what output it obtains. For AAJ, name the weight/Jacobian and BCH ingredients and state the perturbative scope. Include one sentence explaining why these are related constructions rather than four instances of one theorem.

Starred problems.

5*. Full Schmidt rank, not a geometry test. To avoid reusing the regular-coordinate symbol $q$, let $0<\zeta <1$ and

$$|{\Psi }_{\mathrm{TFD}}⟩=\sqrt{\zeta }|00⟩+\sqrt{1-\zeta }|11⟩,$$

and normalize

$$|{\Psi }_{\alpha }⟩={\mathcal{N}}_{\alpha }\left(\sqrt{1-\alpha }|{\Psi }_{\mathrm{TFD}}⟩+\sqrt{\alpha }|00⟩\right),0\le \alpha \le 1.$$

Determine for which $\alpha$ the coefficient matrix has full rank and hence the vector is cyclic and separating for ${\mathcal{A}}_R$. At $\alpha =1/2$, compute its two Schmidt probabilities, then write $\Delta ={\rho }_R\otimes {\rho }_L^{-1}$ and the corresponding $J$. Verify that $J{\mathcal{A}}_{R}J={\mathcal{A}}_R^{\prime }$ whenever the vector has full Schmidt rank. Why does this calculation supply no sharp criterion for a connected bulk geometry?

6*. Where the constant lives. Let ${\hat{\tau }}^{\prime }=c\hat{\tau }$ with $c>0$, and hold the normalized state functional fixed. Show that its density changes from $D$ to $D^{\prime }=D/c$, and therefore

$$S_{{\hat{\tau }}^{\prime }}=-{\hat{\tau }}^{\prime }(D^{\prime }\log D^{\prime })=S_{\hat{\tau }}+\log c.$$

Why does this not affect entropy differences computed with one fixed trace normalization?

7*. Liu’s three branches. For each of the following data, say what Liu’s §VIII.3 proposal concludes and which extra hypothesis is doing work: (a) both boundary algebras are type I; (b) both are type III${}_1$ and the bulk is classical; (c) neither is type I and the bulk is quantum volatile. Then explain why knowing only $J{\mathcal{A}}_{R}J={\mathcal{A}}_R^{\prime }$ answers none of the three classification questions.

8*. Classical or volatile? Consider three hypothetical $G_N\to 0$ families: (i) all invariant fluctuations are $O(G_N^{1/2})$ and all invariant lengths are $O(1)$; (ii) fluctuations are $O(G_N^{1/2})$ but one proper length is $O(G_N^{-1})$; (iii) all invariant sizes are $O(1)$ but the variance of an invariant area remains $O(1)$. Classify each family using Liu’s definition. State why the classification is not determined by algebra type alone.

Project problems.

9. Final-write-up draft due. Submit the ≥10-page draft (begun Wk 9). Incorporate the course research map to position your topic relative to Witten/CPW/AAJ. Feedback only; no grade.

10. Source-scope audit. Build a four-column table for Liu §§V, VII, VIII.C, and IX: exact algebraic input; bulk assumption; limiting procedure; conclusion. Add one final row for AAJ and record that its series uses unitary covariance, a deformation-dependent weight/Jacobian, and BCH. The table must not label AAJ’s method as a Connes-cocycle expansion or equate its entropy coefficients with a bulk time advance without a separate calculation.

9. Instructor checkpoints (internal)

1. Dual action: $e^{irQ}\lambda (t)e^{-irQ}=e^{itr}\lambda (t)$ and $e^{irQ}$ fixes $\pi (a)$. Fourier-mode invariance leaves only the zero mode $\pi (\mathcal{A})$; the second crossing, not the fixed-point statement alone, gives $\mathcal{A}\bar{\otimes }\mathcal{B}(L^2\mathbb{R})$.
2. Exponential profile: normalization is immediate, $H=1-\log \eta$, $\mathbb{E}[p]=1/\eta$, and $S_{\hat{\tau }}=S(\rho )+1-\log \eta -1/\eta$. The mean term comes from the trace weight; the first two distribution terms depend on ${\mu }_{\eta }$.
3. Gaussian derivatives: ${\partial }_{\sigma }S=1/\sigma$ and ${\partial }_{p_0}S=-1$. Neither differentiates a spacetime cutoff.
4. Research map: Witten must be labeled perturbative canonical, CPW strict microcanonical, AAJ weight/Jacobian plus BCH through the stated orders, and CLPW an observer construction whose positive-energy compression is essential for finiteness.
5. Schmidt rank: the $|11⟩$ coefficient is nonzero exactly for $\alpha <1$, so standardness fails only at $\alpha =1$. At $\alpha =1/2$, the Schmidt probabilities are $(1+\sqrt{\zeta })/2$ and $(1-\sqrt{\zeta })/2$. In the Schmidt basis $J$ swaps the factors and conjugates coefficients; $\Delta ={\rho }_R\otimes {\rho }_L^{-1}$.
6. Trace constant: $D^{\prime }=D/c$ and the shift is $+\log c$; it cancels from differences taken with one normalization.
7. Three branches: type I/type I gives disconnected; type III${}_1$/type III${}_1$ plus a classical bulk gives a classical wormhole; non-type-I algebras plus volatility gives a quantum wormhole. These are branches of Liu’s conditional proposal, not consequences of the standard-pair identity.
8. Volatility: family (i) is classical; (ii) is volatile because an invariant length grows as a negative power of $G_N$; (iii) is volatile because invariant fluctuations do not vanish.
9. Final draft rubric: every use of “if and only if” must list the pure semiclassical bulk, no-baby-universe, $G_N\to 0$, and ${\alpha }^{\prime }\to 0$ assumptions.
10. Source audit: §V supplies crossed-product structure, §VII subregion/subalgebra and island context, §VIII.3 the conditional connectivity proposal, and §IX observer dressing and quantum-volatile applications. AAJ is a later comparison, not a theorem stated in those Liu sections.

---

Notes prepared for the algebraic-QFT course, Semester II Block 3. Last revised 2026-09-29.