Bulk Entanglement and Quantum Corrections

The first quantum correction to holographic entropy is the entropy of bulk quantum fields across the classical RT/HRT surface, combined with the renormalized geometric couplings that make the sum finite. This Faulkner–Lewkowycz–Maldacena (FLM) term is of order $N^0$ when the classical area term is of order $N^2$. Neither term is separately regulator independent.

Required background. Quantum Information supplies von Neumann entropy and reduced bulk states, while Gauge Theory and String-Theory Preparation supplies the large-N loop expansion and holographic dictionary. Helpful background. Quantum Gravity supplies gravitational effective-field-theory renormalization, and Black-Hole Information supplies the general entropy and code-subspace setting.

One-loop entropy across the classical surface

Fix a boundary region $A$, a semiclassical bulk state, and its classical extremal surface $\gamma_A$. Let $r_A$ be the bulk region bounded by $A\cup\gamma_A$. In Einstein gravity coupled to $O(1)$ light fields, the entropy through one-loop order has the schematic form

$$\begin{aligned} S_{\mathrm{CFT}}(A)={}& \frac{A_0(\gamma_A)+\delta_g A(\gamma_A)}{4G_{\rm ren}} +S_{\rm bulk,ren}(r_A)\\ &+S_{\rm local}^{(1)}(\gamma_A)+O(G_N). \end{aligned}$$

$A_0$ is computed in the leading metric, while $\delta_g A$ is its change from the order-$G_N$ metric backreaction sourced by the quantum stress tensor. Thus $\delta_g A/(4G_N)$ contributes already at order one. It is included implicitly if “area” is evaluated in the semiclassically corrected metric. $S_{\rm bulk}$ is the von Neumann entropy of the regulated bulk-field algebra in $r_A$ relative to its complement. $S_{\rm local}^{(1)}$ collects the remaining Wald-like, edge, and local counterterm contributions in a fixed algebra and subtraction prescription, without counting them again in $S_{\rm bulk,ren}$. The complete first correction, including backreaction, appears in Faulkner, Lewkowycz, and Maldacena 2013, § 2.2, Eq. (2.15).

The formula is valid in a code subspace whose excitations do not change the leading saddle or overwhelm the $1/N$ expansion. It assumes a common state preparation and regulator on the replica geometries. A number of light species scaling like $1/G_N$ would promote the bulk entropy to leading order and invalidate the stated counting.

Replica derivation of the FLM structure

For integer $n$, compute the normalized boundary Rényi trace by a bulk path integral on a replica saddle. Near $n=1$, write its effective action as

$$I_n-nI_1=(n-1)\frac{\operatorname{Area}(\gamma_A)}{4G_{\rm bare}} +W_n-nW_1+O((n-1)^2),$$

where $W_n$ is the one-loop effective action of fluctuations on the replicated geometry. The von Neumann entropy follows from

$$S(A)=\left.(n\partial_n-1)I_n\right|_{n=1}.$$

The conical response of the classical action produces the area term. The replica derivative of $W_n$ contains the bulk entropy, the response of the background to quantum stress, and local geometric terms. Local divergences renormalize $G_N$ and higher-curvature entropy couplings. This separation is given by Faulkner, Lewkowycz, and Maldacena 2013, § 2.2, Eqs. (2.8)–(2.15).

Why can the entropy term be evaluated on the unshifted classical surface at this order? This is a change of the surface embedding, distinct from the metric backreaction above. Write $X=\gamma_A+\eta$, with $\eta=O(G_N)$. Since $\gamma_A$ is classically extremal,

$$\delta\operatorname{Area}[\gamma_A]=0,$$

so its linear displacement contributes nothing to the order-$N^0$ area. The quadratic area change scales as $\eta^2/G_N=O(G_N)$, while the change of an $O(1)$ bulk entropy under $\eta$ is also $O(G_N)$. Thus the first correction uses $S_{\rm bulk}$ on $\gamma_A$; the displacement becomes essential at the next order and leads naturally to the Quantum Extremal Surface condition.

Renormalization is a cancellation in the sum

Entanglement entropy across a sharp surface has local ultraviolet divergences. In $D$ bulk dimensions its leading form is proportional to the area of the cut,

$$S_{\rm bulk,bare} =c_{D-2}\frac{\operatorname{Area}(\gamma_A)}{\epsilon^{D-2}} +c_{D-4}\frac{1}{\epsilon^{D-4}} \int_{\gamma_A}{\cal I}_2+\cdots .$$

The same loop diagrams renormalize $1/G_N$ and higher-curvature coefficients. Consequently a cutoff variation that raises $S_{\rm bulk,bare}$ shifts the bare geometric entropy oppositely, leaving the generalized sum fixed. The relation between entanglement divergences and gravitational-coupling renormalization is analyzed explicitly by Susskind and Uglum 1994, §§ II–IV, pp. 2702–2709. A calculation that quotes a finite bulk entropy obtained in one subtraction scheme but combines it with an area coupling from another scheme has no invariant meaning.

This cancellation is local but state dependence is not. The divergent coefficients depend on the field content and local geometry; the finite bulk entropy depends on correlations throughout $r_A$. A code-subspace change can alter the latter even when the classical metric and leading area agree. That is precisely why the order-$N^0$ term contains information absent from the classical surface alone.

Checks and validity limits

Classical limit. If $G_N S_{\rm bulk}$ and all one-loop local terms vanish while the geometry is held fixed, the expression reduces to RT/HRT. This checks the order counting but not the regulator cancellation.

Pure-state complement check. For a globally pure regulated bulk state and a common factorization prescription, the entropies on the two sides of $\gamma_A$ agree. Combined with the same area, this reproduces $S(A)=S(\bar A)$ when the boundary state is pure and the same branch applies.

Replica and heat-kernel check. In AdS3 quotients, direct one-loop determinants can be compared with bulk entanglement across the geodesic; such calculations provide an independent realization of the FLM term Barrella et al. 2013, §§ 2–4. Agreement depends on matching boundary conditions and zero-mode treatment.

The derivation does not establish a nonperturbative factorization of gravitational Hilbert space. Gauge constraints create centers and edge terms; replica symmetry can fail; competing classical saddles can become degenerate; and a topology change can invalidate expansion about one $\gamma_A$. Near such a transition, all branches must be treated in the same scheme before dominance is decided.

Exercises

Derive the surface-shift order. Suppose the area near a nondegenerate classical extremum is $A(x)=A_0+\kappa x^2/2$ and the finite bulk entropy is $S_{\rm bulk}(x)=s_0+s_1x$. Find the stationary point of $A/(4G_N)+S_{\rm bulk}$ and determine the orders of its shift and entropy change.

Solution

Differentiating gives $\kappa x/(4G_N)+s_1=0$, hence $x_*=-4G_Ns_1/\kappa$. The displacement is therefore first order in $G_N$ when the Hessian and entropy derivative are order one. Substitution gives $S_{\rm gen}(x_*)=A_0/(4G_N)+s_0-2G_Ns_1^2/\kappa$. Thus the shift changes the generalized entropy only at order $G_N$, one order beyond the FLM term $s_0$. The conclusion fails when $\kappa$ has a zero mode or $s_1$ scales inversely with $G_N$.

References

• Barrella, Taylor, Xi Dong, Sean A. Hartnoll, and Victoria L. Martin. “Holographic Entanglement beyond Classical Gravity.” Journal of High Energy Physics 09 (2013): 109. DOI. Open PDF.
• Faulkner, Thomas, Aitor Lewkowycz, and Juan Maldacena. “Quantum Corrections to Holographic Entanglement Entropy.” Journal of High Energy Physics 11 (2013): 074. DOI. Open PDF.
• Susskind, Leonard, and John Uglum. “Black Hole Entropy in Canonical Quantum Gravity and Superstring Theory.” Physical Review D 50 (1994): 2700–2711. DOI. Open PDF.