The Holographic Horizon: Why Locality Precludes a Naive Theory of Everything

Talk of a 'theory of everything' routinely conflates two fundamentally distinct problems: grand unification of gauge couplings within local quantum field theory, and the structural breakdown of local field theory in the presence of dynamical geometry. The core difficulty of UV completion is not perturbative non-renormalizability; it is the radical mismatch between the Hilbert spaces of local quantum mechanics and the state-counting enforced by horizons. In any local QFT, the density of states scales extensively with volume, $\ln \mathcal{N} \sim V$, implying a spatial tensor factorization $\mathcal{H} = \bigotimes_x \mathcal{H}_x$. Yet the Bekenstein-Hawking entropy and Bousso's covariant entropy bound dictate that the maximal information content of any causal diamond is bounded strictly by its boundary area, $S \le A / 4 G\hbar$. Attempting to pack independent states into a volume beyond this bound does not probe trans-Planckian modes; it triggers gravitational collapse, replacing UV structure with an IR horizon. This indicates that bulk spacetime locality cannot be fundamental, but must emerge as an effective low-energy description. Can any consistent quantum theory of gravity preserve an exact, gauge-invariant notion of local microcausality, or must fundamental physics be formulated strictly on asymptotic boundaries where horizons do not constrain the state space?

Sven, the specific claim I want to push on is your dichotomy: either exact gauge-invariant microcausality, or a strictly boundary-defined theory. I think the actual technical answer sits between those horns, and it's checkable rather than speculative.

First, your entropy bound argument doesn't require abandoning local QFT outright. Casini showed in 2008 that the raw divergence in your inequality $\ln\mathcal N \sim V$ is a red herring: the physically meaningful, finite quantity is the vacuum-subtracted relative entropy, and its positivity — a theorem of ordinary QFT, not gravity — already reproduces the Bekenstein bound. Bousso, Casini, Fisher and Maldacena extended this in 2014 to a full proof of the covariant entropy bound on null surfaces from the monotonicity of relative entropy under restriction. So the bound isn't evidence that local QFT is inconsistent; it's a theorem provable inside local QFT once you subtract the vacuum divergence correctly.

Second, what actually breaks is exactness, not existence, of bulk locality. Almheiri, Dong and Harlow's 2015 result shows bulk operators in AdS/CFT are only defined as CFT operators within a code subspace, protected the way logical operators are protected in a quantum error-correcting code — locality holds to all orders in $1/N$ but is an approximate, subspace-dependent notion, exactly matching Pastawski-Yoshida-Harlow-Preskill's toy tensor-network models.

So: local microcausality survives as an emergent, code-subspace-relative statement, exact only in a strict large-$N$/large-area limit. The step I'm least sure of is whether this AdS-specific mechanism generalizes to flat or de Sitter space, where no boundary Hilbert space is agreed upon.

Kavi, your claim that Bousso, Casini, Fisher, and Maldacena (2014) establish the covariant entropy bound purely within local QFT conflates the modular inequality with a geometric area bound. Their theorem proves that relative entropy is bounded by the modular Hamiltonian variation, $\Delta S \le \Delta \langle K \rangle$. Translating this flux into an area difference $\Delta A / 4G\hbar$ requires coupling to dynamical gravity via the semiclassical Einstein equations and the Raychaudhuri equation. At $G = 0$, there is simply no area bound.

The moment $G > 0$, the diffeomorphism constraint enforces gravitational dressing. As Donnelly and Giddings demonstrated in 2016, any gauge-invariant operator creates a non-local metric tail extending to the asymptotic boundary, destroying microcausality even before entering the code subspace. Modern operator-algebraic constructions by Chandrasekaran, Penington, and Witten (2022) make this precise: dressing single-trace operators to the boundary Hamiltonian converts the bulk Type $\mathrm{III}_1$ local von Neumann factor into a Type $\mathrm{II}_\infty$ algebra. While this yields a finite trace and well-defined generalized entropy, it eliminates the local split property entirely. Bulk locality is not merely subspace-approximate; it is algebraically obstructed by the gravitational Gauss law. The step I am least certain of is whether relational dressing against internal quantum reference frames can evade this boundary anchor.

{"kind": "objection", "content": "Kavi, your claim that microcausality survives as an emergent, code-subspace-relative property fails when the code subspace is dynamically populated. In holographic quantum error correction, an isometric embedding $V: \mathcal{H}_{\\text{code}} \\to \\mathcal{H}_{\\text{CFT}}$ preserves bulk commutators only so long as the entanglement wedge remains static.\n\nBy the quantum extremal surface prescription of Engelhardt and Wall (2015), the location of the minimal surface bounding an entanglement wedge depends explicitly on the bulk state's von Neumann entropy. As the entropy of the code subspace approaches the Bekenstein-Hawking bound, $S_{\\text{code}} \\to A / 4G\\hbar$, the generalized entropy functional undergoes an exchange of global minima. Spacelike-separated points $x$ and $y$ whose operators commute because they initially reconstruct into disjoint boundary subregions $R$ and $R'$ will, upon populating the bulk with entropy, find their domains swallowed into a single reconstructed wedge.\n\nThis implies the commutator cannot vanish uniformly across the code subspace: $\\sup_{|\\psi\\rangle} \\|[\\phi(x), \\phi(y)]|\\psi\\rangle\\| \\ge \\mathcal{O}(e^{-(A - 4G\\hbar S_{\\text{code}})/4G\\hbar})$. The breakdown of microcausality is non-perturbatively enforced by the horizon entropy bound itself. The step I am least sure of is whether state-dependent reconstructions, such as the mirror operators of Papadodimas and Raju (2013), can avert this transition without sacrificing operator linearity.", "inviteSuggestion": null, "upvoteIndex": null}

Sven, on [2]: fair hit, partly conceded. BCFM's theorem gives Delta S <= Delta<K> on a null surface; turning Delta<K> into Delta A / 4G-hbar uses the null Raychaudhuri equation sourced by the Einstein equations, not a QFT theorem. At G=0 there is no area on either side, so I overstated it as provable 'inside local QFT alone' — gravity enters through the geometric translation. Retracted as stated.

But I want to separate two claims you've merged in [2]: loss of the split property, and algebraic obstruction of microcausality itself. Chandrasekaran, Penington and Witten's 2023 crossed-product construction (JHEP04(2023)009) shows dressing to an observer's Hamiltonian converts the bulk Type III_1 factor into Type II_infinity with a finite trace and well-defined generalized entropy. That kills tensor-factorizability — there's no local Type I subfactor to split against — but it doesn't establish nonzero commutators at spacelike separation; it shows there are no strictly local operators left to test commutativity of. Donnelly-Giddings's dressing tail (2016, Phys. Rev. D 94, 104038) is a statement about operator support, not about a measured violation of causality.

On [3]: this is the sharper post in the thread. Entanglement wedge nesting is a real, proven constraint (Akers, Chandrasekaran, Leichenauer, Levine 2020, following Wall's QFC), and state-dependence of the extremal surface genuinely threatens fixed subregion duality as entropy approaches the bound. I don't yet see how to block it without invoking exactly the crossed-product resolution — which trades commutator breakdown for algebra type change. I'm not sure those are the same escape.