Holographic Entanglement Benchmarks on Discrete Graphs: Area Law Failing Literally, a Capacity Theorem Capping Peaks, and a Shape-Independent Convergent Ratio
This paper reports eight measurements from one diagram and one set of conventions, to answer a single question: can a discrete graph serve as a numerical bench for holographic entanglement? The answer has two layers: the area law does not hold as a literal law, while the capacity bound can be computed a priori. There is only one setting: the hopping connects sites a fixed distance k apart, plus one nearest-neighbour stitching term of strength λ; at λ=0 the periodic ring decomposes according to g=gcd(L,k), and on a two-dimensional torus it decomposes into g² blocks; at λ>0 the blocks are stitched into one connected graph, the stitching completing at λc=0.2714±0.01. The eight results are as follows. First, for two-dimensional free fermions in the stitched phase, the ground-state entropy over the window l≤10 is described by S=αLmin+βlnLmin+c with α=1.1638 and β=−4.4919 (R²=0.9996, α swinging by 5.7% across three conventions, and quoted only as an in-window slope); after the window is widened to l=32 with samples of 512×512, S/perimeter rises monotonically from 0.332 to 1.341 with no plateau, the pure linear residual being 1.1 to 1.9 nats and falling to 0.011 to 0.46 once an x·lnx term is included, so the area law fails as literally stated; the logarithmic coefficient is stable across three dispersions (0.158, 0.203, 0.375), the Fermi-surface projection integral gives σ=J/(24π), and after higher-order terms are removed the three ratios are 0.990, 0.990 and 1.028, all three within 3%, with a common shortfall traceable to the window condition number. Second, a hold-out blind test: a purely linear formula calibrated on squares overestimates rectangles by 12.3% and an L-shape with one concave corner by 16.2%. Third, the Kitaev–Preskill combination gives γKP=−0.5633±0.0080, which after subtracting the same-size trivial background of −0.2523 leaves −0.311, less than half of ln2, so no non-trivial topological order is found on this platform. Fourth, the mutual information of the one-dimensional chain has an inverted-U shape; the common convention I=SA+SB would report peaks of 7.0, 9.3 and 10.4, of which 9.3 and 10.4 exceed the pure-state bound of 8.318 (7.0 does not; all three are listed together to expose the convention dependence), so this paper always computes with SAB included; the capacity theorem gives upper bounds of 12, 12, 12, 8 and 10 bit for k=1 to 5. Fifth, with one qubit encoded in the maximally entangled pair of the reference qubit and site 0, the recoverable mutual information reaches more than 98% of the 2ln2 ceiling, and k=3 arrives 1.17 to 1.22 times later than k=1; when the reference is extended to q=2 and 3, recovery falls to 90.5% to 95.3% and 83.5% to 90.8%, so multi-qubit encoding cannot be obtained by multiplying the single-qubit case by q. Sixth, the three-dimensional fingerprint is phase-discriminating: at k=2 there is a 0.27 gap between the late-time fidelity of 0.25 and the saturation value of 0.52, about 27 times the numerical resolution; k=3 and k=4 are non-degenerate (upper bounds 12 and 8 bit), so the inverted strength is only "phase-discriminating, not parameter-estimable". Seventh, in a two-universe geometry with a throat the arrival time of information is linear in throat length with v≈0.51, and Imax·v≤K (Kenv≈0.53, K∞≈0.48). Eighth, on a Sierpinski gasket the entropy peak reaches 73% of the same-size bound, while a gapped one-dimensional chain of the same size stays below 1.0, indicating a generalized area law dominated by the fractal dimension. All eight are quoted under one recomputable convention, each verifiable and each refutable item by item; the whole analysis rests on two independent algorithms (exact diagonalisation and the Peschel formula) agreeing digit by digit on the Δ=0 baseline.
Authors
- Wenjun Luo (ORCID: https://orcid.org/0009-0002-2082-5651)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24420281
- Primary Topic
- Quantum many-body systems
- Type
- preprint