Sem II Week 4 — Rindler Crossed Product: Exact Checks and Honest Limits

The purpose of a mini-calculation is to expose the mechanism, not to decorate an abstract theorem with a formal integral. We therefore divide the calculation into three layers. The Rindler wedge gives an exact type III${}_1$ algebra and exact geometric modular flow. The inner-action Fourier model gives an exact trace, density, and entropy calculation. A regulated free field gives the familiar matter area law. These layers fit together, but none can substitute for the others. In particular, the width of the crossed-product clock is not the matter UV cutoff, and a same-clock Weyl excitation has exactly the same trace entropy as the vacuum because it is an inner unitary conjugation.

0. Reading

Primary:

• Witten, “Gravity and the crossed product,” arXiv:2112.12828, §§3.3–3.5. Section 4 concerns other conserved charges; it is not a free-field section.
• Sem I Weeks 13–14 and Appendix D.

Free-field background:

• Bisognano and Wichmann, J. Math. Phys. 16 (1975) 985 and 17 (1976) 303.
• Bombelli, Koul, Lee, Sorkin, Phys. Rev. D 34 (1986) 373.
• Srednicki, Phys. Rev. Lett. 71 (1993) 666.

0.1 How to use this master dossier

• Classroom core: §§1–2, 3.1–3.3, 4.1–4.2, and 5.1–5.3, with Problems 1–5. This route puts one exact theorem, one exact trace model, and one exact cancellation on the board.
• Full derivation / self-study: §§3.4, 4.3, 5.4, and 6, with Problems 6–10. This route audits units, trace normalization, non-inner comparisons, and the independent matter regulator.
• Research extension: §§7–8 and Problems 11–12. This route asks which conclusions survive in Witten’s gravitational setting and which require genuinely new bulk input.

The chapter is intentionally layered. A student should always be able to label a displayed equation as an exact wedge-algebra statement, an exact inner-action control calculation, a regulated QFT statement, or a gravitational comparison.

1. Scope of the calculation

1.1 Exact QFT input

We use the course’s mostly-plus convention ${\eta }_{\mu \nu }=\mathrm{diag}(-1,+1,+1,+1)$. For the right Rindler wedge

$$W_R=\{x\in {\mathbb{R}}^{1,3}:x^1>|x^0|\},$$

let $\mathcal{A}(W_R)$ be the vacuum wedge algebra of a free scalar field. Under the usual hypotheses it is the hyperfinite type III${}_1$ factor. Bisognano–Wichmann gives

$${\Delta }_{W_R}=e^{-2\pi K_{\mathrm{boost}}},{\sigma }_t(a)=e^{i2\pi t{K}_{\mathrm{boost}}}a{e}^{-i2\pi t{K}_{\mathrm{boost}}}.$$

Therefore

$$\hat{\mathcal{A}}(W_R):=\mathcal{A}(W_R){⋊}_{\sigma }\mathbb{R}$$

is a type II${}_{\infty }$ factor with a faithful normal semifinite trace. These statements are exact.

For orientation, take the real Klein–Gordon field with

$$(\square -m^2)\varphi =0,\square =-{\partial }_0^2+{\nabla }^2,$$

and form Weyl operators $W(f)=e^{i\varphi (f)}$ from real test functions modulo the field equation. With $E$ the causal propagator,

$$W(f)W(g)=e^{-iE(f,g)/2}W(f+g).$$

If $\Lambda (u)$ is the boost in the $(x^0,x^1)$ plane and $f_u(x):=f(\Lambda (-u)x)$, covariance gives

$$e^{iu{K}_{\mathrm{boost}}}W(f)e^{-iu{K}_{\mathrm{boost}}}=W(f_u).$$

The displayed Bisognano–Wichmann convention therefore says ${\sigma }_t(W(f))=W(f_{2\pi t})$. Writing the pullback explicitly is useful: it fixes the sign in $f_{-2\pi q}$ below instead of leaving the sign to memory.

1.2 What is not being claimed

There is no ordinary density matrix for the vacuum restricted to $\mathcal{A}(W_R)$. Nor will we derive the canonical trace by inserting nonnormalizable vectors $|0_M⟩\otimes |{\delta }_q⟩$. Finally, the free scalar does not contain Newton’s constant, so it cannot derive the coefficient $1/(4G_N)$.

The calculation instead checks:

1. the crossed-product covariance and all convention signs;
2. the canonical trace scaling in an exact Fourier model;
3. the density and entropy of explicit product states in that model;
4. unitary invariance for a local coherent excitation;
5. the independent origin of the matter UV area law.

1.3 Variable dictionary

Symbol	Meaning
$t$	dimensionless modular parameter
$u=2\pi t$	physical boost rapidity
$q$	scalar coordinate in the regular representation
$Q$	multiplication operator, $(Q\xi )(q)=q\xi (q)$
$P=-i{\partial }_q$	conjugate momentum; $\lambda (t)=e^{-itP}$
$p$	spectral variable of $P$ after Fourier transform
$r$	dual-action parameter

Calling $P$ “position” or treating $g(P)$ as multiplication by $g(q)$ would invalidate the trace calculation.

2. Sub-task 1: construct the Rindler core

On $\mathcal{F}\otimes L^2({\mathbb{R}}_q)$ define

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

Since the modular action is geometric,

$${\sigma }_{-q}(W(f))=W(f_{-2\pi q}),$$

where the subscript of $f_{-2\pi q}$ is a physical rapidity. This distinction matters: $q$ is the regular-representation coordinate, while $-2\pi q$ is the rapidity selected by the modular parameter $-q$.

The covariance check is

$$\begin{array}{rl}(\lambda (t)\pi (W(f))\lambda (t{)}^{\ast }\xi )(q)& ={\sigma }_{-(q-t)}(W(f))\xi (q)\\ & =(\pi ({\sigma }_t(W(f)))\xi )(q).\end{array}$$

Thus

$$\lambda (t)\pi (W(f))\lambda (t{)}^{\ast }=\pi ({\sigma }_t(W(f))).$$

The dual action is

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

implemented in the ambient algebra by $e^{irQ}$. In the $P$-spectrum it is $p\mapsto p-r$.

3. Sub-task 2: exact trace calculation in the Fourier model

3.1 Why we change models

For the type III wedge, the existence of $\hat{\tau }$ is an abstract theorem. To compute a weighted integral without hiding domain questions, replace the wedge algebra temporarily by $M_n(\mathbb{C})$ with a faithful density matrix $\rho$. The modular action is inner, so the crossed product can be untwisted and Fourier transformed:

$$M_n(\mathbb{C}){⋊}_{{\sigma }^{\rho }}\mathbb{R}\cong M_n(\mathbb{C})\bar{\otimes }{L}^{\infty }({\mathbb{R}}_p).$$

This is a type-I algebra with centre; it is not the Rindler core. It is nevertheless the right laboratory for the trace formula.

The unitary equivalence can be seen directly. In the regular representation,

$$({\pi }_{\rho }(a)\xi )(q)={\rho }^{iq}a{\rho }^{-iq}\xi (q).$$

Define the decomposable unitary

$$(V\xi )(q)={\rho }^{-iq}\xi (q).$$

Then

$$V{\pi }_{\rho }(a)V^{\ast }=a\otimes 1,V\lambda (t)V^{\ast }={\rho }^{-it}\otimes \lambda (t).$$

Because ${\rho }^{it}\in M_n(\mathbb{C})$, the algebra generated by $a\otimes 1$ and ${\rho }^{-it}\otimes \lambda (t)$ is the same as the algebra generated by $M_n(\mathbb{C})\otimes 1$ and $1\otimes \lambda (t)$. Fourier transformation sends the abelian group von Neumann algebra generated by $\lambda (t)$ to $L^{\infty }({\mathbb{R}}_p)$. This derivation explains both features of the model: the crossed product has a centre because the original action was inner, and $p$—not $q$—is the multiplication variable in the final trace formula.

3.2 A genuine trace ideal

Let $a\ge 0$ be a matrix and let $f\ge 0$ be bounded with compact support. Then

$$\hat{\tau }(a\otimes f(P))=\mathrm{Tr}(a){\int }_{\mathbb{R}}f(p)e^{-p}dp<\infty .$$

Finite linear combinations of such positive elements generate a concrete $\hat{\tau }$-finite ideal. No distributional vectors and no unspecified “regularization” are required.

3.3 Dual scaling

Because ${\theta }_r(f(P))=f(P-r)$,

$$\begin{array}{rl}\hat{\tau }({\theta }_r(a\otimes f(P)))& =\mathrm{Tr}(a)\int f(p-r)e^{-p}dp\\ & =e^{-r}\mathrm{Tr}(a)\int f(p^{\prime })e^{-p^{\prime }}d{p}^{\prime }\\ & =e^{-r}\hat{\tau }(a\otimes f(P)).\end{array}$$

This proves the scaling law in the model with the same normalization used for the type III core.

3.4 Rindler normalization

The variable $p$ belongs to the auxiliary Fourier model. To compare normalizations, define $E:=p/(2\pi )$; if a physical construction identifies this auxiliary spectral coordinate with a boost-energy collective variable, then

$$e^{-p}dp=2\pi e^{-2\pi E}dE.$$

The constant $2\pi$ is absorbed into the arbitrary normalization of the type II${}_{\infty }$ trace. This reproduces the form of the Unruh weight in the comparison variable $E$, not a derivation that the auxiliary $p$ equals the matter boost generator. In either case it is not a weight $e^{-2\pi q}$ on the regular coordinate $q$.

4. Sub-task 3: density and entropy of an explicit state

4.1 Product state

Choose a probability density $\mu (p)$ and define

$$\hat{\omega }(a\otimes f(P))=\mathrm{Tr}(\rho a)\int \mu (p)f(p)dp.$$

Relative to $\hat{\tau }$, its density is

$$D(p)=\rho \mu (p)e^p.$$

Indeed,

$$\hat{\tau }(D(a\otimes f(P)))=\mathrm{Tr}(\rho a)\int \mu (p)f(p)dp.$$

4.2 Entropy

Assume $S(\rho )<\infty$, $\int \mu |\log \mu |<\infty$, and $\int |p|\mu (p)dp<\infty$. Since the matrix and abelian factors commute,

$$\log D(p)=\log \rho +\log \mu (p)+p.$$

The trace weight cancels the density factor before the logarithm is averaged:

$$\begin{array}{rl}{S}_{\hat{\tau }}(\hat{\omega })& =-\int e^{-p}\mathrm{Tr}\left[\rho \mu (p)e^p(\log \rho +\log \mu (p)+p)\right]dp\\ & =-\mathrm{Tr}(\rho \log \rho )-\int \mu (p)\log \mu (p)dp-\int p\mu (p)dp.\end{array}$$

Therefore

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

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

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

The mean term comes from the nonuniform trace measure; it must be retained.

If the displayed integrability assumptions fail, the entropy can be extended-real or undefined through an $\infty -\infty$ subtraction. The algebra supplies a trace; it does not guarantee that every normal state’s entropy is a finite real number.

4.3 Trace rescaling audit

For ${\hat{\tau }}^{\prime }=c\hat{\tau }$, the same state has density $D^{\prime }=D/c$, and

$$S_{{\hat{\tau }}^{\prime }}=S_{\hat{\tau }}+\log c.$$

This sign agrees with Witten and CPW after translating their trace convention.

5. Sub-task 4: coherent-state cancellation

5.1 The state is an inner conjugate

Let $f$ be real and supported in $W_R$. Then $W(f)\in \mathcal{A}(W_R)$, and its image $U=\pi (W(f))$ is a unitary in the core. If the vacuum lift has density $D_0$ and the coherent state uses the same clock and the same core identification, then

$$D_f=U{D}_0{U}^{\ast }.$$

5.2 Entropy is invariant

Functional calculus gives $\log D_f=U(\log D_0)U^{\ast }$, so

$$\begin{array}{rl}{S}_{\hat{\tau }}(D_f)& =-\hat{\tau }(U{D}_0\log D_0{U}^{\ast })\\ & =-\hat{\tau }(D_0\log D_0)=S_{\hat{\tau }}(D_0).\end{array}$$

Hence

$$\boxed{S_{\hat{\tau }}(D_f)-S_{\hat{\tau }}(D_0)=0.}$$

5.3 Relative entropy and modular energy cancel

Applying the exact entropy-difference identity,

$$0=-S(D_f\Vert D_0)+\hat{\tau }[(D_f-D_0)K_0],K_0=-\log D_0.$$

Thus

$$\boxed{S(D_f\Vert D_0)=\Delta ⟨K_0⟩.}$$

This is the correct coherent-state result for an inner Weyl excitation. A formula with two nonzero terms is not wrong if the terms cancel; it is wrong if it presents their sum as an independent nonzero entropy change.

5.4 When a nonzero answer is possible

A nonzero result can arise if the excitation is not inner relative to the chosen algebra, if the clock distribution changes, or if two state-dependent cores are compared through a nontrivial cocycle identification. The calculation must say which mechanism is present.

6. Sub-task 5: the matter area law

6.1 The regulated free-field entropy

The matter vacuum entropy is computed in a regulated type-I approximation to the local QFT. In $3+1$ dimensions,

$$S_{\mathrm{matter}}(W_R;ϵ)=c_2\frac{A_{\perp }}{{ϵ}^2}+c_{\log }\log \frac{L}{ϵ}+\cdots ,$$

where $c_2$ is regulator-dependent. For the infinite Rindler plane it is more natural to quote entropy per unit transverse area.

A lattice count explains the power without pretending to determine the coefficient. Discretize each transverse direction with spacing $ϵ$. A patch of entangling surface of area $A_{\perp }$ meets approximately

$$N_{\perp }\sim \frac{A_{\perp }}{{ϵ}^2}$$

independent near-boundary transverse cells. Local vacuum correlations across the cut give an $O(1)$ entropy per regulated cell, so the leading contribution scales as $N_{\perp }$. The precise $c_2$ depends on the regulator and field content. For a flat plane in a massless conformal theory, a universal logarithmic term need not be present; logarithms arise when masses, curvature, corners, or other scales supply the required dimensionless ratio. The leading four-dimensional statement relevant here is the area law.

In $1+1$ dimensions there is no transverse area. A CFT interval has

$$S_{\mathrm{interval}}=\frac{c}{3}\log \frac{\ell }{ϵ}+\text{const},$$

and a half-line has the corresponding one-boundary coefficient $c/6$ after introducing an infrared scale.

Thus “logarithmic divergence” is the characteristic $1+1$-dimensional statement, whereas “area divided by ${ϵ}^2$” is the leading $3+1$-dimensional statement. Writing one formula for both dimensions hides the geometry that the divergence is measuring.

6.2 Why this is not the clock entropy

For a Gaussian $\mu$, the clock contribution is

$$H(\mu )=\log \sigma +\frac{1}{2}\log (2\pi e),$$

which tends to $-\infty$ as $\sigma \to 0$. The matter entropy tends to $+\infty$ as $ϵ\to 0$. Besides the opposite sign, the four-dimensional matter divergence carries a transverse area. A one-dimensional clock distribution cannot produce that factor.

Therefore

$$\boxed{\sigma \text{and}ϵ\text{are independent scales}.}$$

The crossed product provides a renormalized trace framework; it does not manufacture the matter UV area law.

7. Comparison with Witten’s gravitational construction

Layer	Rindler/QFT statement	Witten statement
Algebra	$\mathcal{A}(W_R)$ is type III${}_1$	strict-$N$ noncentral algebra is type III${}_1$
Modular flow	exact boost by $u=2\pi t$	TFD modular flow generated by ${\beta }_H\hat{H}$
Core	type II${}_{\infty }$ by Takesaki’s structure theorem	perturbative energy-inclusive algebra is type II${}_{\infty }$
Trace	abstract on type III core; exact in Fourier model	explicit on Witten’s classical-quantum trace domain
Entropy	exact semifinite identity	black-hole entropy up to a constant for the specified states
UV law	independent regulated matter calculation	gravitational area term fixed by the semiclassical dictionary

The free-field construction verifies the algebraic mechanism and the convention signs. It does not verify Witten’s gravitational identification in a theory without gravity.

8. What is proved, modeled, and imported

• Proved in the QFT setup: Bisognano–Wichmann modular flow; type III${}_1$ wedge algebra under the stated hypotheses; type II${}_{\infty }$ continuous core.
• Computed exactly in the inner-action model: trace ideal, dual scaling, product-state density, entropy, and trace-rescaling sign.
• Computed exactly for the coherent lift: unitary invariance and relative-entropy/modular-energy cancellation.
• Imported from regulated QFT: the logarithmic $1+1$ law and the $A/{ϵ}^2$ leading divergence in $3+1$.
• Not derived here: $1/(4G_N)$, the black-hole area response, or the full Witten entropy formula.

9. What to take away

• $q$ is the regular coordinate, $Q$ multiplies by $q$, and $P=-i{\partial }_q$. The trace weight belongs to the Fourier variable $p$, not to $q$.
• The canonical law is $\hat{\tau }\circ {\theta }_r=e^{-r}\hat{\tau }$.
• A valid trace check uses a genuine trace ideal, such as finite-rank matrices tensored with compactly supported functions of $P$.
• Product entropy is $S(\rho )+H(\mu )-\mathbb{E}[p]$.
• Same-clock inner Weyl excitation implies zero entropy difference and exact cancellation between relative entropy and modular energy.
• Clock width and matter cutoff are distinct. The crossed product does not derive the free-field area law.
• The Rindler example tests the algebraic skeleton; gravity supplies the generalized-entropy interpretation.

10. Problem set

Core problems.

1. Covariance. Verify the crossed-product covariance on a Schwartz vector and track the boost pullback convention for $f_{-2\pi q}$.

2. Canonical pair. Show $[Q,P]=i$, $\lambda (t)=e^{-itP}$, and $e^{irQ}P{e}^{-irQ}=P-r$. Deduce the dual action on $f(P)$.

3. Trace ideal. Choose $a\ge 0$ in $M_2(\mathbb{C})$ and $f(p)=1_{[-L,L]}(p)$. Compute $\hat{\tau }(a\otimes f(P))$ and its dual transform explicitly.

4. Gaussian density. Derive $D(p)=\rho \mu (p)e^p$ and $S=S(\rho )+\frac{1}{2}\log (2\pi e{\sigma }^2)-p_0$.

5. Coherent cancellation. Starting from $D_f=U{D}_0{U}^{\ast }$, prove entropy invariance and the equality of relative entropy and modular-energy increase.

6. Two scales. Make a two-column dimensional-analysis table for the distribution width $\sigma$ and the matter regulator $ϵ$. Explain why a scalar width in modular-energy space cannot supply $A_{\perp }/{ϵ}^2$.

Starred problems.

7*. Witten’s trace. Reproduce Witten’s §3.4 trace formula and translate it with $p=-X$. State the analytic or decay assumptions on the operator.

8*. Non-inner comparison. Give a finite-dimensional pair of faithful densities not related by unitary conjugation. Compute $\Delta S$, relative entropy, and the modular-energy term, and verify the exact identity numerically or symbolically.

9*. Regulated Rindler modes. Introduce a transverse lattice cutoff and explain how the number of near-boundary oscillators scales as $A_{\perp }/{ϵ}^2$. Keep this calculation separate from the clock distribution.

10*. Cocycle variant. Describe how a state-dependent core identification can turn a simple product lift into a non-product lift. Identify the point at which the inner-unitary cancellation no longer applies.

Project problems.

11. Block-1 final writeup. Write a 7–10 page account with three explicitly labeled layers: exact Rindler modular theory, exact Fourier trace model, and gravitational interpretation. Any statement crossing from one layer to another must name the extra assumption.

12. Reproducibility appendix. Prepare a one-page convention sheet containing $t,u,q,Q,P,p,r$, the two Fourier conventions, the trace weight, and the entropy-rescaling sign. Another student should be able to reproduce every sign from that page alone.

11. Instructor checkpoints (internal)

1. Covariance: $\lambda (t{)}^{\ast }$ first sends $q$ to $q+t$; the subsequent translation produces ${\sigma }_{-(q-t)}={\sigma }_{-q}{\sigma }_t$. With $f_u(x)=f(\Lambda (-u)x)$, the fibre smearing is $f_{-2\pi q}$ if the rapidity label is written explicitly.
2. Canonical pair: on $C_c^{\infty }$, $[Q,P]=i$; Baker–Campbell–Hausdorff gives $e^{irQ}P{e}^{-irQ}=P-r$, hence ${\theta }_r(f(P))=f(P-r)$.
3. Trace ideal: for $f=1_{[-L,L]}$, $\hat{\tau }(a\otimes f)=\mathrm{Tr}(a)(e^L-e^{-L})$. Its dual transform has support $[r-L,r+L]$ and trace $e^{-r}\mathrm{Tr}(a)(e^L-e^{-L})$.
4. Gaussian: $D=\rho \mu e^p$; the $e^p$ cancels the trace weight. The entropy is $S(\rho )+\frac{1}{2}\log (2\pi e{\sigma }^2)-p_0$.
5. Coherent cancellation: functional calculus gives $\log (U{D}_0{U}^{\ast })=U\log D_0{U}^{\ast }$ and traciality gives $\Delta S=0$. The exact difference identity then forces $S(D_f\Vert D_0)=\Delta ⟨K_0⟩$.
6. Two scales: $\sigma$ is a width in the dimensionless auxiliary $p$ variable; $ϵ$ is a spacetime length cutoff. Only the latter combines with $A_{\perp }$ to give $A_{\perp }/{ϵ}^2$.
7. Witten trace: the accepted derivation uses his analytic strip and contour shift; a diagonal ${\delta }_q$ matrix element receives no credit.
8. Non-inner pair: for commuting faithful matrices with eigenvalues $r_i$ and $s_i$, check $\Delta S=-{\sum }_i{r}_i\log r_i+{\sum }_i{s}_i\log s_i$, $S(r\Vert s)={\sum }_i{r}_i\log (r_i/s_i)$, and $\Delta ⟨K_s⟩=-{\sum }_i(r_i-s_i)\log s_i$. These satisfy the identity term by term.
9. Rindler lattice: the number of transverse cells meeting the cut is $N_{\perp }\sim A_{\perp }/{ϵ}^2$; an $O(1)$ short-distance contribution per cell yields the leading area scaling, not its regulator-dependent coefficient.
10. Cocycle variant: the state-change isomorphism modifies the group implementer by a relative modular cocycle. After this joint transformation the lifted density need not equal $U{D}_0{U}^{\ast }$ for one unitary in one fixed core, so the inner-unitary proof no longer applies.
11. Final-writeup rubric: each cross-layer arrow must name the new input; in particular, the free field never supplies $G_N$.
12. Convention sheet: it must include ${\sigma }_t=\mathrm{Ad}{\Delta }^{-it}$, $\lambda (t)=e^{-itP}$, $p=-X_{\mathrm{W}}$, and $\hat{\tau }\circ {\theta }_r=e^{-r}\hat{\tau }$.

---

Notes prepared for the algebraic-QFT course, Semester II Block 1. Last revised 2026-09-29.

End of Sem II Block 1.