Sem II Week 3 — Trace, Entropy, and the Generalized-Entropy Interpretation

Last week we built Witten’s perturbatively corrected right algebra ${\mathcal{A}}_R={\mathcal{A}}_{R,0}{⋊}_{{\sigma }^{\mathrm{TFD}}}\mathbb{R}$, a type II${}_{\infty }$ factor. We now follow the calculation in §§3.4–3.5 of the paper: construct its trace, identify the density of a classical–quantum state, and compute its entropy. We then put this result beside the black-hole generalized entropy, keeping the logical layers separate. Witten uses the JLMS relation and a choice of gravitational normalization to interpret the answer; CPW develops the more systematic semiclassical generalized-entropy statement. Section 4 of Witten’s paper is about additional conserved charges, not entropy.

Route through this master dossier. Classroom core (two-hour seminar): §§1–4, with the trace, density, entropy formula, and Gaussian check derived on the board. Full derivation or self-study: the analytic details in §§2–3, the interpretation ledger in §5, and Problems 1–5. Research extension or office hour: §§6–7, the starred problems, and the comparison with CPW and JLMS. The short route is a pacing choice, not a shortened version of the subject.

0. Reading

Primary:

• Witten, “Gravity and the crossed product,” arXiv:2112.12828, §§3.1, 3.4–3.5.
• Sem I Wk 14, especially the trace-rescaling discussion and the entropy-difference identity.

Secondary:

• Chandrasekaran, Penington, Witten, “Large $N$ algebras and generalized entropy,” arXiv:2209.10454, entropy sections.
• Jafferis, Lewkowycz, Maldacena, Suh, “Relative entropy equals bulk relative entropy,” arXiv:1512.06431.
• Faulkner, Lewkowycz, Maldacena, “Quantum corrections to holographic entanglement entropy,” arXiv:1307.2892.

1. What is being computed

Let ${\mathcal{A}}_{R,0}$ be the strict-large-$N$ type III${}_1$ factor and let $\Psi$ be the TFD reference vector. In Witten’s representation the crossed product is generated by ${\mathcal{A}}_{R,0}$ together with bounded functions of

$${\hat{h}}_{\Psi }+X,$$

where ${\hat{h}}_{\Psi }=-\log {\Delta }_{\Psi }$ generates the modular flow and $X$ is the collective energy coordinate introduced in Week 2. We denote the resulting algebra by

$${\mathcal{A}}_R={\mathcal{A}}_{R,0}{⋊}_{{\sigma }^{\Psi }}\mathbb{R}.$$

The algebra is type II${}_{\infty }$, so it admits a faithful normal semifinite trace. This statement is operator-algebraic. The interpretation of $X$ as a rescaled black-hole energy fluctuation is the gravitational input.

1.1 A convention warning

Witten uses a coordinate $X$ for which the trace weight is $e^{X}dX$. Other presentations use $-X$, $s=X/(2\pi )$, or a Fourier-dual variable; their weights consequently look like $e^{-X}dX$ or $e^{-2\pi s}ds$. These formulas agree after the declared change of variable. They cannot be combined before that change is made.

2. Witten’s trace

2.1 An analytic core

Consider an element written formally as

$$\hat{a}={\int }_{\mathbb{R}}dua(u)e^{iu({\hat{h}}_{\Psi }+X)},$$

where $a(u)\in {\mathcal{A}}_{R,0}$ decays sufficiently rapidly and extends holomorphically to the KMS strip. On this analytic core, Witten obtains

$$\mathrm{Tr}\hat{a}={\int }_{\mathbb{R}}dX{e}^X⟨\hat{\Psi }|\hat{a}|\hat{\Psi }⟩=2\pi ⟨\Psi |a(i)|\Psi ⟩.$$

The second equality follows by shifting the $u$ contour through the KMS strip and then performing the $X$ integral. The formula is therefore not an ordinary Lebesgue integral for arbitrary bounded operators; the analytic-domain hypotheses are part of the statement.

2.2 Why it is a trace

For two analytic elements $\hat{a}$ and $\hat{b}$, the KMS boundary relation moves one coefficient through the reference state while the contour shift changes its modular-time argument. The result is

$$\mathrm{Tr}(\hat{a}\hat{b})=\mathrm{Tr}(\hat{b}\hat{a}).$$

This calculation is worth reading in Witten §3.4 because it shows exactly where KMS analyticity enters. Semifiniteness and extension from the analytic core to the full crossed product are imported from the continuous-core theorem.

2.3 Trace scaling

The shift $X\mapsto X+c$ is the dual action in this representation. It rescales the trace:

$$\mathrm{Tr}⟼e^c\mathrm{Tr}.$$

If a normalized state has density $\rho$ relative to $\mathrm{Tr}$, the rescaled trace has density $e^{-c}\rho$. Hence

$$S(\rho )=-\mathrm{Tr}(\rho \log \rho )⟼S(\rho )+c.$$

The sign is fixed: multiplying the trace by $e^c$ shifts every normalized-state entropy by $+c$.

3. The classical–quantum state and its density

3.1 State preparation

Choose a probability density $g(X)\ge 0$ with

$${\int }_{\mathbb{R}}g(X)dX=1,$$

and consider

$$\hat{\Psi }=\Psi \otimes g(X{)}^{1/2}.$$

Different functions $g$ describe different states of the collective energy coordinate. Changing $g$ is not a trace renormalization and need not shift all entropies by the same constant.

3.2 Density relative to the trace

Witten factorizes the modular operator of $\hat{\Psi }$ into a crossed-product part and a commutant part. The crossed-product factor is

$$K=e^{-({\hat{h}}_{\Psi }+X)}g({\hat{h}}_{\Psi }+X),$$

possibly understood as an affiliated positive operator if the displayed function is unbounded. It satisfies

$$\mathrm{Tr}(\hat{a}K)=⟨\hat{\Psi }|\hat{a}|\hat{\Psi }⟩,$$

and is therefore the density of the vector state relative to the semifinite trace.

3.3 Entropy, derived

We can compute the entropy without diagonalizing $K$. Since $K$ is the density,

$$S(K)=-\mathrm{Tr}(K\log K)=-⟨\hat{\Psi }|\log K|\hat{\Psi }⟩.$$

Now

$$\log K=-({\hat{h}}_{\Psi }+X)+\log g({\hat{h}}_{\Psi }+X).$$

The reference modular generator annihilates its vector, ${\hat{h}}_{\Psi }\Psi =0$. Evaluating the remaining multiplication operators in the state $g^{1/2}$ gives

$$\boxed{S(K)={\int }_{\mathbb{R}}dX[Xg(X)-g(X)\log g(X)].}$$

Thus the answer has two terms: the mean collective coordinate $⟨X{⟩}_g$ and the differential entropy of its probability density. Neither term is, by itself, the matter entanglement entropy across a sharp QFT boundary.

3.4 Gaussian check

For

$$g(X)=\frac{1}{\sqrt{2\pi {\sigma }^2}}\exp \left[-\frac{(X-X_0{)}^2}{2{\sigma }^2}\right],$$

the integrals are elementary:

$$\int Xg(X)dX=X_0,-\int g\log gdX=\frac{1}{2}\log (2\pi e{\sigma }^2).$$

Therefore

$$\boxed{S(K)=X_0+\frac{1}{2}\log (2\pi e{\sigma }^2).}$$

This is a genuine, reproducible model calculation. It also exposes two common mistakes. First, narrowing the Gaussian sends the differential-entropy term to $-\infty$, not $+\infty$. Second, the clock width $\sigma$ is not a matter UV cutoff and does not reproduce an area law.

4. Where the black-hole area enters

4.1 The JLMS relation used by Witten

Earlier in the paper, Witten reviews the formal one-sided modular-Hamiltonian relation

$${\beta }_H{H}_R+\log Z=h_r+\frac{A}{4G_N},$$

with the analogous formula on the left. Here $h_r$ is the bulk effective-field modular contribution and $A/(4G_N)$ is the area operator contribution. The important point is the combination: the separate terms need not be well-defined operators at the required order, while the boundary modular Hamiltonian is.

Equivalently, after subtracting the thermal expectation value, Witten writes a relation of the form

$${\beta }_{H}NU+⟨S⟩=h_{\ell }+\frac{A}{4G_N}.$$

This relation explains why the collective energy coordinate can carry the leading gravitational entropy after a choice of trace normalization. It does not identify the Shannon entropy of an arbitrary clock profile with $S_{\mathrm{out}}$.

4.2 What is established in Witten 2022

The source supports the following statements:

• the perturbatively corrected algebra is type II${}_{\infty }$;
• its entropy is defined up to a state-independent additive constant;
• the explicit state above has entropy $\int (Xg-g\log g)dX$;
• with the chosen normalization, entropy is measured relative to the classical black-hole entropy, and the Gaussian energy fluctuations produce a universal logarithmic correction.

The stronger formula

$$S_{\mathrm{II}}=\frac{A}{4G_N}+S_{\mathrm{out}}+\text{constant}$$

for a suitable class of semiclassical states belongs to the combined Witten–JLMS–CPW framework. In this course we cite CPW for the systematic generalized-entropy statement rather than attributing its full derivation to Witten §4.

5. Entropy differences: an exact model identity

The following finite-dimensional identity is useful because it separates algebra from holography. Let $\rho$ and $\sigma$ be faithful density matrices and set $K_{\sigma }=-\log \sigma$. From

$$D(\rho \Vert \sigma )=\mathrm{Tr}\rho (\log \rho -\log \sigma ),$$

we obtain exactly

$$\boxed{S(\rho )-S(\sigma )=-D(\rho \Vert \sigma )+\mathrm{Tr}[(\rho -\sigma )K_{\sigma }].}$$

No gravity has entered. If a crossed-product dressing realizes the analogous identity for a specified family of states, the second term is a modular-energy difference. Identifying that term with an area variation requires JLMS or another holographic input; identifying the first term with a bulk relative entropy requires entanglement-wedge reconstruction. These are separate statements.

5.1 Infinitesimal first law

For $\rho =\sigma +\delta \rho$, relative entropy begins at second order, so

$$\delta S=\delta ⟨K_{\sigma }⟩$$

at first order. This is the first law of entanglement. It determines variations around a reference state, not a finite area formula without further input.

6. What the free-field analog can check

The Rindler model provides explicit modular flow by Bisognano–Wichmann and therefore gives a concrete continuous core. It can check:

1. covariance of the crossed-product generators;
2. the dual action and trace scaling;
3. the density and entropy for a specified clock distribution;
4. finite relative-entropy or modular-energy differences for specified states.

It cannot, without additional physics, check:

• the large-$N$ and order-by-order $1/N$ origin of Witten’s collective coordinate and its noncentral dressing; the separate holographic scaling $G_N\sim 1/N^2$ is not derived by the free-field analogue either;
• the coefficient $1/(4G_N)$;
• a matter UV area law from the clock width;
• the full holographic identification with bulk generalized entropy.

There is also a useful invariance test. If two dressed states with the same clock profile differ only by an inner unitary of the type II algebra, their entropies are equal. Any decomposition of the difference into relative-entropy and modular-energy terms must then cancel. A nonzero answer in that situation signals that the state, the dressing prescription, or the reference modular Hamiltonian has been changed somewhere in the calculation.

7. Proof-status ledger

Statement	Status in this course
${\mathcal{A}}_{R,0}{⋊}_{\sigma }\mathbb{R}$ is II${}_{\infty }$	stated continuous-core theorem, after verifying the hypotheses
Witten’s trace on the analytic core is cyclic	derived from KMS analyticity, following §3.4
$K=e^{-(\hat{h}+X)}g(\hat{h}+X)$ is the state density	stated from Witten’s modular-operator factorization; checked by the trace pairing
$S(K)=\int (Xg-g\log g)dX$	derived above
trace rescaling shifts entropy by $+\log c$	derived above
JLMS modular Hamiltonian contains $A/(4G_N)$	imported holographic result
type-II entropy equals $S_{\mathrm{gen}}$ for semiclassical states	stated with Witten, JLMS, and CPW scope
clock width is the matter UV cutoff	false; no such identification is made

8. What to take away

• Witten’s entropy calculation is in §§3.4–3.5; §4 treats other conserved charges.
• The source’s exact model formula is $S(K)=\int (Xg-g\log g)dX$.
• The mean collective coordinate and its probability entropy are distinct contributions.
• Rescaling the trace by $c$ shifts every normalized-state entropy by $+\log c$.
• The area interpretation is a holographic normalization using JLMS and the large-$N$ collective-energy relation; it is not a free-field area-law derivation.
• A Rindler continuous core is a laboratory for the algebra, not a derivation of gravity.

9. Looking ahead

Week 4 should now have a precise task: choose one representation and one clock coordinate, verify the covariance and trace-scaling relations, and compute a density and entropy for a specified $g$. It should not try to manufacture $A/(4G_N)$ from an unspecified clock cutoff. Weeks 5–7 then turn to CPW, where the generalized-entropy comparison is developed for the two-sided semiclassical setting.

10. Problem set

Core problems.

1. Trace-rescaling sign. Let ${\tau }^{\prime }=c\tau$ and let $h$ be the density of a normalized state relative to $\tau$. Show that $h^{\prime }=h/c$ and that $S_{{\tau }^{\prime }}=S_{\tau }+\log c$.

2. Density check. Starting from $\mathrm{Tr}(\hat{a}K)=⟨\hat{\Psi }|\hat{a}|\hat{\Psi }⟩$, verify $\mathrm{Tr}K=1$ and derive the entropy formula in §3.3.

3. Gaussian clock. Reproduce §3.4. Study separately the limits $\sigma \to 0$ and $\sigma \to \infty$, and explain why neither is a matter UV limit.

4. Entropy-difference identity. Derive the boxed identity in §5 for two noncommuting density matrices. Then expand it to first order around $\sigma$ and recover the first law of entanglement.

5. Source map. Read Witten §§3.4–3.5 and locate the trace formula, the density $K$, the entropy formula, and its rescaling. Then inspect §4 and state its actual subject in one sentence.

Starred problems.

6*. JLMS layer. Read the operator relation quoted in Witten §3.1 and JLMS arXiv:1512.06431. Identify which equality is an operator-algebra statement and which step uses holography.

7*. Analytic-core trace. For $\hat{a}=\int dua(u)e^{iu(\hat{h}+X)}$, reproduce the contour-shift argument giving $\mathrm{Tr}\hat{a}=2\pi ⟨\Psi |a(i)|\Psi ⟩$. List the decay and strip-analyticity assumptions you used.

8*. Inner-unitary check. In a finite semifinite factor, let ${\rho }_U=U\rho U^{\ast }$. Prove $S({\rho }_U)=S(\rho )$ and show explicitly how the relative-entropy and modular-energy terms cancel in the identity of §5.

Project problem.

9. Witten-to-CPW comparison. Write a four-page account distinguishing: (i) Witten’s explicit entropy calculation; (ii) the JLMS area relation; (iii) CPW’s semiclassical generalized-entropy theorem; and (iv) the free-field continuous-core check. Every paragraph should state which layer it belongs to.

---

Notes prepared for the algebraic-QFT course, Semester II Block 1. Last revised 2026-08-24.