Entropy, Geometry, and Motives:

Interfaces Between Quantum Gravity,

Information, and Thermodynamics
HAMID JAVANBAKHT
DBA Sebastian Ruliad, Isotelesis Inc.
[email protected], Mountain View, CA

August 25, 2025

Abstract

We study how entropy and geometry interact across quantum gravity, information theory,
non-equilibrium thermodynamics, coding theory, and arithmetic/categorical geometry. On the
gravity side we organize the Ryu–Takayanagi area law together with replica-based, quantum, and
higher-derivative corrections, and relate boundary terms and junction conditions to interface cal-
culus. On the information side we review statistical distance (Fisher/Bures/Fubini–Study), dual
connections, and generalized entropies as sensitivity functionals. On the coding/combinatorics
side we connect holographic tensor-network models and quantum LDPC codes to expansion in
Ramanujan graphs/complexes. On the categorical side we summarize six-functor formalisms
and relative non-Archimedean geometry as algebraic backbones for gluing.
Our contributions are: (i) a comparative concept map and dependency graph spanning
these areas; (ii) a set of reusable stencils—boundary/interface primacy, variational principles,
metric–area/cut dictionaries, expansion/coding templates, and categorical gluing—with explicit
knobs and toggles; and (iii) a menu of falsifiable tests (e.g., turning on/off bulk-entanglement or
higher-curvature terms; varying expander parameters; near-equilibrium vs. far-from-equilibrium
regimes). We state scope conditions throughout (e.g., linear response limits; distinctions between
classical area, bulk entropy, and higher-derivative functionals; existence vs. efficient reconstruc-
tion). The result is a structured foundation for a unified treatment of entropy, geometry, and
motives.

1
Contents
1 Introduction — Why entropy, why geometry, why now? 3

2 Information Geometry and Quantum Statistical Distance 4

3 Gravitational Entropy, Interfaces, and Variational Structure 6

4 Entanglement, Quantum Corrections, and Error Correction 8

5 Arithmetic & Categorical Backbones:

Motives and Non-Archimedean Geometry 10

6 Shape Optimization, Interfaces, and PDE Variations 12

7 Cross-Domain Synthesis: Common Stencils and Productive Tensions 15

8 Outlook and Open Problems 18

9 Section→Reference Concordance 20

A Glossary 23

B Acronym Index 29

C Symbol Index 30

D Experiment Matrix & Falsifiers 34

2
1 Introduction — Why entropy, why geometry, why now?
Motivation and scope. Entropy is a counting principle, a metric of distinguishability, and
a variational functional. It links quantum gravity (black-hole mechanics and holographic entan-
glement), information theory (statistical distance, estimation bounds), non-equilibrium thermody-
namics (production and reciprocity), coding/combinatorics (expansion and redundancy), and arith-
metic/categorical geometry (gluing and realization). Our aim is not to advocate a single definition
of entropy, but to extract structural stencils that recur across these domains: boundary/interface
primacy, extremal principles, metric/area dictionaries, expansion-driven robustness, and categorical
gluing. These stencils will be made explicit and reusable, with knobs (model choices) and toggles
(on/off tests) that enable falsification.

Historical baselines and generalizations. Classical traditions anchor the story: Boltzmann–
Gibbs entropy as a state count; Shannon entropy as an information measure; and the thermody-
namic laws as constraints on transformations. Two one-parameter families broaden this baseline.
Rényi entropies interpolate sensitivities to tails via power sums, isolating a tunable emphasis on
common versus rare events. Tsallis entropies relax additivity, modeling systems with long-range
interactions or constraints that deviate from ideal mixing. On the dynamical side, near-equilibrium
thermodynamics organizes flux–force pairs, entropy production, Curie symmetry reductions, and
Onsager–Casimir reciprocity, providing a precise regime where linear response and convex pro-
duction principles hold. Together these generalizations supply a toolkit for choosing the “right”
sensitivity and the “right” regime for a given phenomenon.

Information geometry as method. Information geometry recasts statistical models as Rie-

mannian manifolds. The Fisher information defines the canonical metric; dual affine connections
(∇, ∇∗ ) and their α-family encode complementary coordinate systems (mixture vs. exponential),
projection theorems, and Bregman divergences. In quantum settings, state spaces inherit metrics
from Hilbert-space structure: for pure states, the Fubini–Study line element matches statistical
distinguishability; for mixed states, the quantum Fisher/Bures geometry bounds estimation via
the quantum Cramér–Rao inequality. Curvature can be read as fluctuation content on coherent-
state manifolds. This geometric language turns “how much can we tell two states apart?” into
differential statements that we will later compare with geometric areas and minimal cuts.

Interfaces, boundaries, and the categorical turn. A recurring motif in gravity, analysis, and
algebra is that interfaces carry law. In Euclidean gravity, well-posed actions require boundary terms
and regularity conditions; in relativistic matching, junction conditions encode extrinsic-curvature
jumps across thin shells; in PDE design, Hadamard structure theorems concentrate shape gradients
on boundaries. Categorical frameworks offer an abstract mirror of this interface logic. The 20th-
century shift from sets to sheaves, topoi, and spectra reframed “space” as what glues, culminating in
motivic ideas where cohomology theories are organized by functorial six-operations and localization

3
triangles. This historical/categorical vignette motivates our use of motivic backbones in tandem
with physical interfaces.

Why holography, codes, and expanders appear. Holography elevates boundary/interface

logic: entanglement entropies of boundary subregions are captured by bulk extremal surfaces.
Quantum error-correcting codes provide operational models of bulk–boundary encoding, while high-
dimensional expanders (and Ramanujan complexes) furnish explicit discrete geometries with strong
mixing and redundancy—attributes that stabilize reconstruction against erasures. These combina-
torial/coding structures let us test “area vs. cut” correspondences and probe existence vs. efficient
reconstruction, a tension sharpened by computational-complexity constraints.

Beyond Shannon/Rényi/Tsallis: motivic directions. On the arithmetic/categorical side,

motivic entropy frameworks attach information quantities to exponentiable motivic measures and
track information loss along functorial maps. In parallel, non-Archimedean integral geometry
(Cauchy–Crofton-type formulas) and Vitushkin-style invariants offer complexity/entropy proxies
adapted to Berkovich spaces. These developments suggest arithmetic analogs of our metric/area
dictionary and point toward “boundary objects” within premotivic and six-functor contexts.

Contributions and structure. We provide (i) uniformly structured per-file reports (definitions,
results, knobs/toggles, reuse hooks); (ii) a cross-domain synthesis that isolates common stencils
and productive tensions; and (iii) falsifiable toggle-style tests (e.g., add/remove bulk-entropy terms
or higher-derivative corrections; vary expander parameters; switch boundary conditions in shape
calculus; propagate entropy along categorical functors). Section 2 reviews information-geometric
foundations. Section 3 develops gravitational entropy and interface calculus, feeding Section 4 on
entanglement, quantum corrections, and error-correcting structures. Section 5 assembles arith-
metic/categorical backbones (A1 -homotopy primer, six-functor formalism, motivic entropy, and
non-Archimedean kinematics). Section 6 records shape/variation tools. Section 7 synthesizes sten-
cils and tensions; Section 8 outlines open problems and tests.

2 Information Geometry and Quantum Statistical Distance

Statistical manifolds and the Fisher metric. Let {pθ (x)}θ∈Θ⊂Rd be a regular parametric
family (smooth in θ, identifiable, dominated). The Fisher information

gij (θ) = Eθ ∂i log pθ (X) ∂j log pθ (X) = − Eθ ∂i ∂j log pθ (X) (1)

   

defines a Riemannian metric on Θ. It quantifies local distinguishability and appears in the (classical)
Cramér–Rao bound Varθ (θ̂) ⪰ g(θ)−1 upon unbiasedness/regularity assumptions. This metric is
invariant under sufficient-statistic reduction and under smooth reparametrizations, making it the
canonical geometry for inference.

4
Dual connections, α-geometry, and projections. Information geometry equips (Θ, g) with
a pair of torsion-free affine connections (∇, ∇∗ ) that are dual with respect to g (i.e. ∂k gij = Γkij +
Γ∗kji in suitable coordinates). The α-family ∇(α) interpolates between the mixture (α = −1)
and exponential (α = +1) geometries, with (∇(α) )∗ = ∇(−α) . Exponential families are ∇(1) -
flat in natural parameters; mixture families are ∇(−1) -flat in expectation parameters. For convex
potentials φ and Legendre dual φ∗ , the Bregman divergence

Dφ (p∥q) = φ(p) − φ(q) − ⟨∇φ(q), p − q⟩.

Dφ (p∥q) = φ(p) − φ(q) − ⟨∇φ(q), p − q⟩ (2)

induces orthogonal projection theorems and Pythagorean relations (the e- and m-projections).
These projection identities underwrite algorithms such as iterative scaling and EM-type proce-
dures, and they provide a reusable “orthogonality” stencil we will mirror later in bulk/boundary
reconstructions.

Quantum statistical distance: pure and mixed states. For pure states |ψ⟩ on a Hilbert
space, the Fubini–Study metric is

ds2 = ⟨dψ| dψ⟩ − |⟨ψ| dψ⟩|2 , (3)

the natural Riemannian structure on projective space P(H). For mixed states ρθ , the symmet-
ric logarithmic derivative (SLD) Li solves ∂i ρ = 2 (Li ρ
1
+ ρLi ), and the (SLD) quantum Fisher
information (QFI)
Q
(θ) = 1
Tr ρθ {Li , Lj } (4)


gij 2

−1
yields the quantum Cramér–Rao bound Cov(θ̂) ⪰ g Q (θ) for locally unbiased estimators, achiev-
able under suitable measurement conditions. The corresponding Riemannian metric generates the
Bures distance dBures (ρ, σ) and coincides with the Helstrom metric; for unitary families it reduces
to the variance of the generator (a curvature–fluctuation link). On manifolds of generalized co-
herent states, curvature quantities directly encode fluctuation data, giving concrete, computable
geometries.

Generalized entropies and induced geometries. Beyond the Shannon/Kullback–Leibler

P set-
1− p(x)q
ting, Rényi entropies Hα (p) = 1
log α and Tsallis entropies Sq (p) = provide
P
x p(x)
x
1−α q−1
tunable sensitivities to tails and concentration. These choices select different divergence fam-
ilies and, consequently, different geodesics and projection rules on the same statistical model.
Operationally, the choice of divergence/geometry acts as a knob that trades off robustness and
sensitivity—a theme we will reprise when we compare area-only functionals to higher-curvature
and bulk-entropy corrections in gravity.

5
Interfaces to the rest of the paper. (1) Metric ↔ area/cut dictionary: statistical distance
and QFI act as “infinitesimal distinguishability,” which we later compare to geometric extremals
(areas/min-cuts) in holography and codes. (2) Duality ↔ reciprocity: (∇, ∇∗ ) mirrors Onsager–
Casimir pairs in near-equilibrium thermodynamics and adjoint functors in categorical gluing. (3)
Projections ↔ reconstruction: e/m-projections provide an abstract template for boundary-to-bulk
reconstruction rules and decoding maps.

Knobs, toggles, and caveats.

• Knobs: divergence choice (KL/Bregman vs. α-divergences), connection parameter α, coor-

dinate system (mixture vs. exponential), measurement model (POVM) in the quantum case.

• Toggles: pure vs. mixed states; classical vs. quantum metrics; finite- vs. infinite-dimensional
families; regular vs. singular models (boundaries, non-identifiability).

• Caveats: not all monotone quantum metrics coincide; task-dependent optimality must be
stated. In non-regular regimes, standard projection theorems can fail and require tailored
geometry.

Takeaway. Information geometry turns inference and distinguishability into Riemannian data
(metrics, connections, projections). These geometric primitives will serve as the “sensitivity layer”
we align with gravitational entropies (area/functional extremals), coding cuts/thresholds (expanders
and LDPC structure), and categorical gluing (six functors and motivic measures) in the sections
that follow.

3 Gravitational Entropy, Interfaces, and Variational Structure

Euclidean action, boundary terms, and thermodynamics. In the semiclassical (Euclidean)
approach, the gravitational free energy is obtained from the on–shell action

1
Z
√ 1
Z √
IE [g] = − g (R − 2Λ) dd x − h K dd−1 x + Ict , (5)
16πGN M 8πGN ∂M

where the Gibbons–Hawking–York boundary term makes the variational problem well-posed under
Dirichlet boundary data and Ict denotes counterterms (e.g. in asymptotically AdS spacetimes).
Demanding smoothness at a Euclidean horizon fixes the imaginary-time periodicity, yielding the
Hawking temperature T = κ/(2π) from the surface gravity κ. Thermodynamic quantities then
follow from Z ∼ e−IE , with S = (1 − β∂β ) log Z evaluated on the saddle.

Junctions and thin shells as geometric interfaces. Across a non-null hypersurface Σ with
induced metric hab and unit normal nµ , the extrinsic curvature Kab may jump when a thin shell of

6
stress Sab is present. The Israel junction conditions read

= 8πGN Sab , [X] ≡ X (+) |Σ − X (−) |Σ , (6)

   
Kab − hab K

providing a coordinate–invariant interface law. They supply a template for “boundary–carried”

dynamics that we will mirror in shape calculus and categorical gluing.

Holographic area law (Ryu–Takayanagi). For a boundary region A in a static holographic

state, the entanglement entropy is given at leading order by the area of a bulk minimal surface γA
homologous to A:
area(γA )
SA = . (7)
4 GN
This functional identifies geometry as an extremal record of boundary correlations and sets the
baseline for corrections.

Replica method and extremality (gravitational derivation). Consider the replicated par-
tition function Zn on an n-fold branched cover. The von Neumann entropy follows from
 
SA = 1 − n ∂n log Zn . (8)
n=1

In gravity, this corresponds to saddles with a codimension-2 conical defect of opening 2π/n along
a bulk surface homologous to A. Extremizing the action with respect to the placement of this
“cosmic brane” enforces the extremal (in static cases, minimal) surface condition, reproducing (7)
and fixing the prescription beyond cases with explicit U (1) symmetry.

Higher-derivative corrections (beyond Einstein gravity). For a diffeomorphism-invariant

Lagrangian density L(gµν , Rµνρσ , ∇R, . . .), the entropy is not captured by area alone. A general
functional includes Wald’s Noether-charge term plus extrinsic-curvature corrections:
Z √ h ∂L i
SA = 2π det h εµν ερσ + (extrinsic curvature terms) dd−2 y , (9)
γA ∂Rµνρσ

where εµν is the binormal to γA . The extra terms vanish in Einstein gravity and in special embed-
dings, but are required in generic higher-curvature theories (and are fixed by the replica/extremality
analysis).

Computational inaccessibility as a physical constraint. Thought experiments that would

extract detailed interior information from Hawking radiation often require decoding operations
whose gate complexity scales exponentially with the black-hole entropy. This places otherwise–
kinematically allowed protocols beyond any feasible observer, suggesting a computational refinement
of effective field theory: physical accessibility is limited not only by locality and causality, but also
by complexity barriers. We will contrast this limitation with constructive coding models in §4.

7
Synthesis. This section exhibits three reusable stencils: (i) boundary/interface primacy (GH
boundary term; Israel junctions; extremal surfaces); (ii) variational structure (saddles and replicas
select geometric extremals that encode entropy); and (iii) corrections and regimes (quantum/bulk
contributions and higher-derivative terms refine the area law). Together they frame how geom-
etry encodes information and which knobs (Lagrangian, state, boundary conditions) control the
encoding.

Knobs, toggles, caveats.

• Knobs: gravitational Lagrangian (Einstein vs. higher curvature); cosmological constant;

ensemble/boundary conditions; choice of region A; topology of γA .

• Toggles: classical vs. one-loop/bulk entanglement; smooth vs. singular saddles; time-reflection
symmetry on/off (minimal vs. extremal); thin shell on/off (Israel).

• Caveats: replica analytic continuation is an assumption; null hypersurfaces require a refined

junction formalism; counterterms are scheme–dependent; cosmic-brane derivations assume
controlled backreaction.

4 Entanglement, Quantum Corrections, and Error Correction

Classical RT baseline. In static holographic states, the leading entanglement entropy of a
boundary region A is given by the area of a homologous bulk minimal surface γA :

area(γA )
S(A) = .
4 GN

The homology constraint enforces a consistent “cut” and guarantees subadditivity; nesting of min-
imal surfaces underlies strong subadditivity in generic configurations. Tripartite information is
typically nonpositive (monogamy) in this classical limit, reflecting geometric inequalities on unions
of minimal surfaces.

Generalized gravitational entropy (replica/cosmic brane). The replica method computes

S(A) = (1 − n∂n ) log Zn |n=1 by considering bulk saddles with a codimension-2 defect homologous
to A. The brane’s position is fixed by extremizing the action; in time-reflection symmetric states
this reduces to a minimal surface. This derivation removes the need for an explicit U (1) symmetry
and cleanly identifies the extremality condition as the geometric core of the prescription.

Quantum and one-loop corrections (FLM). Beyond the classical area, bulk fields entangled
across a Cauchy slice ΣA contribute an additional term,

area(γA )
S(A) = + Sbulk (ΣA ) + (counterterms) . (10)
4 GN

8
Extremizing the generalized entropy Sgen = area /(4GN ) + Sbulk + · · · selects a quantum extremal
surface (QES) rather than a purely minimal one, and the entanglement wedge bounded by the QES
replaces the classical wedge. This refines reconstruction statements and explains state-dependent
shifts of γA due to bulk matter entanglement.

Higher-derivative gravity (Dong functional). When the bulk Lagrangian contains higher-
curvature terms, the entropy functional is no longer pure area. Wald’s Noether-charge contribution
appears, supplemented by extrinsic-curvature corrections fixed by the replica/extremality analysis.
These additions vanish in Einstein gravity but are generically required, and they constrain consistent
higher-curvature couplings via entropy inequalities.

Holographic codes and operator reconstruction. Tensor-network models realize a bulk→boundary

isometry V : Hbulk → Hbdy with built-in erasure correction: bulk operators supported inside the
(quantum) entanglement wedge of A can be represented on HA alone. Perfect tensors arranged on
hyperbolic tilings reproduce key geometric features: (i) minimal cuts through the network count
boundary-entropy exactly in idealized settings, and (ii) “operator pushing” implements comple-
mentary recovery on overlapping boundary regions. These models supply constructive proofs-of-
principle for wedge reconstruction and make explicit the role of graph geometry.

Quantum LDPC and high-dimensional expanders. Modern quantum LDPC code families
achieve almost linear minimum distance with bounded-weight checks. Their Tanner graphs and
chain complexes are built from high-dimensional expanders (e.g., Ramanujan complexes), whose
spectral and coboundary expansion enforce large code distance and small logical support. This
bridges the gap between toy holographic codes and scalable fault tolerance: expansion plays the
role of “discrete negative curvature,” while local constraints maintain implementability.

Theme: from areas to cuts to decoders. Equation (10) elevates Sgen to the optimized objec-
tive in gravity; minimal (or quantum extremal) surfaces in the continuum correspond to minimal
cuts in tensor networks. In coding language, minimal cuts lower-bound recoverable logical informa-
tion and relate to distance/threshold. Expansion controls the gap between typical and worst-case
reconstructions, aligning geometric extremality with combinatorial robustness.

Interfaces to other sections. (i) With §3, the replica/QES picture supplies the variational
principle behind entropy functionals and clarifies the status of higher-derivative and bulk-matter
corrections. (ii) With §5, categorical gluing suggests algebraic analogs of bulk/boundary recon-
struction (six functors as “adjoints for codes”). (iii) With §6, Hadamard boundary forms model
how small deformations of γA or network geometry shift Sgen and minimal cuts.

Knobs, toggles, caveats.

9
 • Knobs (gravity): choice of state (thermal/pure), region topology, higher-curvature cou-
plings, matter content, renormalization scheme for counterterms.

• Knobs (codes): tensor choice (stabilizer vs. random), graph/complex degree, code rate vs.
distance, check weight, decoder class.

• Toggles: classical vs. quantum (turn Sbulk on/off); Einstein vs. higher-derivative; connected
vs. disconnected regions; presence/absence of expansion.

• Caveats: exact RT equality holds only in idealized network geometries; continuum AdS/CFT
includes backreaction and 1/N effects; efficient decoders may not exist even when reconstruc-
tion exists (complexity constraints).

5 Arithmetic & Categorical Backbones:

Motives and Non-Archimedean Geometry
5.0 Primer: A1 -Homotopy and Premotivic Intuitions
From spaces to sheaves and spectra. Let Sm/k be smooth k-schemes with a Grothendieck
topology (Nisnevich/étale). Presheaves of simplicial sets on Sm/k become sheaves after descent;
imposing A1 -invariance (contracts along the affine line) and stabilizing by suspension with P1 (or
S 1 ∧ Gm ) yields the motivic stable homotopy category SH(k).

Transfers and premotivic structure. Additive correspondences give sheaves with transfers,
enabling functoriality not only along morphisms but also along finite correspondences. The resulting
“premotivic” formalism supports base change, localization triangles, and purity, anticipating the
six-functor calculus and triangulated motives.

Takeaway. Motivic homotopy recasts algebraic geometry in a homotopical language where gluing
(descent), invariance (A1 ), and stabilization (spectra) are the guiding operations. We will use this
backbone to mirror the boundary/interface logic seen in gravity and PDE shape calculus.

Six-functor formalism, localization, duality

Six operations and adjunctions. For f : X → Y (reasonable schemes), triangulated motives
DM(X) admit
f ∗ ⊣ f∗ , f! ⊣ f ! , − ⊗ − ⊣ Hom(−, −),

with base-change and projection-formula compatibilities. These functors implement the motivic
analogs of restriction, (co)homological pushforward, and exceptional functoriality.

10
Localization and purity. For a closed immersion i : Z ,→ X with open complement j : U ,→ X,
one has distinguished triangles (localization)

+1 +1
j! j ! → id → i∗ i∗ −−→, i! i! → id → j∗ j ∗ −−→,

encoding gluing across an interface. Purity identifies i! with i∗ up to twists/shifts under regularity
hypotheses, providing a clean “boundary condition” dictionary. Duality and traces follow from the
closed monoidal structure.

Realizations and coefficients. Realization functors (Betti, ℓ-adic, de Rham) transport motives
to familiar (co)homology theories compatibly with the six operations. Choices of coefficients (Z vs.
Q), topologies, and transfers act as knobs trading generality for comparison strength.

Why here. The six-functor calculus is the categorical mirror of our boundary/interface theme:
the triangles above are the algebraic counterpart of Israel’s junctions and Hadamard boundary
forms, while adjunctions echo reciprocity structures.

Motivic Entropy & Information Loss

Exponentiable motivic measures. A motivic measure is a ring homomorphism µ : K0 (Vark ) →
R from the Grothendieck ring of varieties to a commutative target R. Exponentiable measures inter-
act well with symmetric powers (via λ-ring/Witt-vector structures), enabling zeta-type generating
series and “additive-to-multiplicative” lifts.

Entropy and loss channels. Given an exponentiable µ, one defines entropy-like functionals and
information loss along two basic channels: (i) coefficient change R → R′ (ring maps deform the
valuation of complexity); (ii) geometric maps f : X → Y (pushforward/pullback along morphisms
reorganize µ-mass). This yields KL-type divergences and functoriality constraints that mirror
Khinchin-style axioms in the motivic setting.

Functoriality via Witt/λ-rings. The Witt-vector (or λ-ring) formalism packages symmetric
powers and zeta constructions, making “entropy of a measure” compatible with products and
filtrations. Kernels/ideals in R identify loci of “least information,” giving a categorical handle on
loss and coarse-graining.

Why here. Motivic entropy extends the Rényi/Tsallis menu to the categorical side: it specifies
how information behaves under the very operations ( f ∗ , f∗ , f! , f ! ) we use to glue and cut. This is
the arithmetic analog of the metric/area and cut-based pictures from earlier sections.

11
Motivic Integral Geometry & Metric Entropy
Cauchy–Crofton in the non-Archimedean world. In classical integral geometry, Crofton
formulas recover measures (length/area) by averaging over incidences with linear subspaces. In
Berkovich/non-Archimedean settings one replaces Euclidean averaging by definable families and
motivic (constructible) measures, obtaining Crofton-type expressions valued in motivic rings.

Vitushkin-style invariants and complexity. Vitushkin invariants V0 , V1 , . . . measure geo-

metric complexity (e.g., covering/oscillation counts). Their motivic versions provide entropy-like
proxies adapted to non-Archimedean geometry, computable via valuations on K0 and compatible
with symmetric-power structures.

Why here. These constructions supply an arithmetic counterpart of our “metric/entropy dictio-
nary”: where Fisher/Bures and RT-area guide the analytic side, motivic Crofton/Vitushkin provide
definable kinematics and complexity on the arithmetic side.

Bridge notes.

• Interfaces/gluing. Localization triangles in §5 implement algebraic “interfaces” analogous

to Israel junctions and shape-calculus boundaries: j! j ! → id → i∗ i∗ is the categorical stencil
of “cut–paste with boundary data.”

• Entropy flows. Motivic entropy/loss (§5) tracks how information moves along f ∗ , f∗ , f! , f ! ;
this mirrors e/m-projections in information geometry and boundary-to-bulk reconstruction
in holography/codes.

• Kinematics/metrics. Non-Archimedean Crofton and motivic Vitushkin invariants (§5)

are arithmetic analogs of lengths/areas and cut capacities, suggesting cross-checks between
analytic and motivic “geodesy.”

• Prospects: boundary motives. One may seek “boundary objects” inside DM(X) encoding
interface data (trace/dual pairs), providing a categorical stage for generalized second laws and
reconstruction constraints.

6 Shape Optimization, Interfaces, and PDE Variations

Speed method, material & shape derivatives. Let Ωt = (id + tV )(Ω) be a C 1 deformation
of a reference domain Ω ⊂ Rd driven by a sufficiently smooth velocity field V . For a state ut
(solution of a PDE on Ωt ) and a functional J(Ωt , ut ), the material derivative u̇ is the Lagrangian
time derivative along particle paths, while the shape derivative

d
dJ(Ω)[V ] = J(Ωt , ut )
dt t=0

12
measures first-order sensitivity to boundary motion. Transport identities,

d d
Z Z Z Z
ft dx = f˙t + ft ∇· V dx, gt ds = ġt + gt ∇τ · Vτ + Hgt Vn ds,



dt Ωt Ωt dt ∂Ωt ∂Ωt

(with mean curvature H, tangential divergence ∇τ · (·), and normal speed Vn = V · n) provide the
bookkeeping for domain/boundary variations.

Hadamard structure theorem (boundary-only gradients). For a broad class of well-posed

PDE-constrained functionals one has the boundary representation
Z
dJ(Ω)[V ] = G[Ω; u, p] Vn ds, (11)
∂Ω

where G is the shape gradient density depending on the state u and an adjoint p. Thus only the
normal component of V matters; tangential deformations reparametrize but do not change J at
first order. Equation (11) is the PDE analog of “interface-carried law” and mirrors our gravitational
and categorical interface stencils.

Adjoint method (prototype: Poisson). Consider

−∇· (κ ∇u) = f in Ω, u = 0 on ∂Ω,

and a tracking cost J(Ω, u) = 2 Ω (u

1R
− ud )2 dx. Introduce the Lagrangian
Z
L(Ω, u, p) = J(Ω, u) + p − ∇· (κ∇u) − f dx.


Ω

Stationarity w.r.t. u gives the adjoint

−∇· (κ ∇p) = u − ud in Ω, p = 0 on ∂Ω,

and the Hadamard density becomes

1 κ
G = (u − ud )2 − κ ∂n u ∂n p + |∇u|2 ,
2 2

up to conventional symmetrizations and boundary-condition terms. Gradient flows Vn = −G (or

quasi-Newton/geodesic updates on shape manifolds) descent J while preserving mesh quality.

Interfaces and transmission conditions. For composite media with an internal interface Γ ⊂
Ω across which coefficients jump (e.g., κ± ), the state satisfies transmission conditions

JuKΓ = 0, Jκ ∂n uKΓ = 0,

13
with J·K the jump across Γ. The shape derivative localizes on both ∂Ω and Γ,
Z Z
dJ(Ω)[V ] = G∂Ω Vn ds + GΓ Vn ds,
∂Ω Γ

where GΓ involves jumps of energy densities and normal fluxes. This is the PDE counterpart
of Israel’s junction conditions: extrinsic geometry (here, normal variations) and flux continuity
together constrain admissible interface motion.

Reciprocity and near-equilibrium templates. Linear irreversible thermodynamics writes

flux–force pairs as Ji = Lij Xj , with Onsager–Casimir reciprocity Lij = ±Lji under time-
P
j
reversal signatures. In steady transport/dispersion design, these relations furnish quadratic pro-
duction functionals whose first variations yield boundary-only shape gradients, aligning with (11).
Thus reciprocity plays the same role here as dual connections (∇, ∇∗ ) in information geometry and
adjunctions (f! , f ! ) in motives.

Curvature terms and geometric regularization. The shape Hessian is typically ill-posed;
regularization adds curvature penalizations Jreg (Ω) = η 2 ds or perimeter terms η Per(Ω),
R
∂Ω H
trading fidelity for stability. Variational updates then couple physics-informed G with geometric
smoothing, analogous to counterterms or higher-derivative corrections in gravity.

Examples and patterns.

• Compliance/thermal compliance. For elasticity or heat, G reduces to energy-like densities

minus flux products; optimal shapes balance stored energy and load work.

• Inversion/identification. With J = 2 Ω |u − ud | ,
1R 2 adjoint-based G drives inclusion bound-
aries Γ toward data-consistent configurations, paralleling wedge reconstruction via minimal
cuts.

• Constrained designs. Isoperimetric, volume, or perimeter constraints add boundary mul-

tipliers, mirroring gravitational ensembles or coding-rate constraints.

Knobs, toggles, caveats.

• Knobs: boundary conditions (Dirichlet/Neumann/Robin); coefficient laws (anisotropy, non-

linearity); regularization weights; geometric constraints (volume/perimeter).

• Toggles: single vs. multi-physics coupling (e.g., thermoelasticity); smooth boundary vs.
corners; stationary vs. unsteady transport.

• Caveats: shape derivatives assume sufficient regularity; topological changes require enrich-
ment (topological derivatives or level-set nucleation); discretization and remeshing introduce
bias and must be controlled (mesh-independent gradients).

14
Interfaces to other sections. (i) The boundary-only representation (11) instantiates our in-
terface primacy stencil next to GH boundary terms and motivic localization triangles. (ii) Trans-
mission conditions echo Israel’s laws at thin shells. (iii) Adjoint reciprocity parallels (∇, ∇∗ ) and
(f! , f ! ). (iv) Curvature/perimeter regularizers are the PDE analog of higher-derivative/counterterm
corrections, and gradient flows of J mirror minimal-surface deformations in §4.

7 Cross-Domain Synthesis: Common Stencils and Productive Ten-

sions
Concept map (one page in words). Four families of objects recur across our corpus: (i) inter-
faces/boundaries (GH boundary terms; Israel junctions; Hadamard forms; categorical localization
triangles); (ii) variational functionals (Euclidean actions, Sgen , PDE costs with boundary-only
gradients); (iii) metrics/entropies (Fisher/Bures/FS; RT area and its higher-derivative/quantum
corrections; motivic entropy and motivic Vitushkin invariants); (iv) reconstruction mechanisms
(bulk–boundary decoding via QECCs; e/m-projections in information geometry; adjunctions/six
functors). These assemble into a reusable dictionary connecting geometry, information, thermody-
namics, codes, and motives.

A working dictionary (objects and roles).

Theme Analytic/physical side Categorical/combinatorial side

Interface carrier GH boundary term; Israel junction Localization triangles j! j ! → id →
[Kab ]; Hadamard density G Vn i∗ i∗ ; adjunctions (f ∗ , f∗ ), (f! , f ! )
Variational
principle Saddle/replica selecting (quantum) Shape derivatives concentrate on
extremal surfaces; Sgen extremiza- ∂Ω and internal Γ; functorial
tion “costs” along six operations
Metric/entropy Fisher/Bures/FS; area/curvature; Motivic entropy & information loss;
bulk entanglement and higher- non-Archimedean Crofton; motivic
curvature corrections Vitushkin V0
Reconstruction Entanglement wedge reconstruc- e/m-projections (IG); functorial
tion; tensor-network cuts; QLDPC pushforward/pullback across covers
decoding and inclusions
Robustness
resource Negative curvature (AdS); spec- High-dimensional expanders (Ra-
tral gap/expansion; reciprocity near manujan complexes); λ/Witt func-
equilibrium toriality (stability under symmetric
powers)

15
Unifying stencils.

• Boundary/interface primacy. Laws concentrate where data glue or change: GH boundary

action; Israel junctions; Hadamard boundary-only gradients; motivic localization triangles.
Interfaces encode the admissible variations and constraints.

• Variational structure. Replica/Euclidean saddles extremize Sgen ; PDE costs yield Hadamard
densities; both produce Euler–Lagrange-type interface conditions. In motives, six-functor cal-
culus supplies the categorical variational “shape” via triangles and adjunctions.

• Metric/entropy dictionary. Distinguishability metrics (Fisher/Bures/FS) correspond to

geometric (area/curvature) or combinatorial (cut/expansion) measures; motivic entropy and
V0 are arithmetic analogs of sensitivity/complexity.

• Dualities and reconstruction. (∇, ∇∗ ) in information geometry parallels (f! , f ! ) in motives

and complementary recovery in holography/codes; reconstruction is an adjoint story across
domains.

• Expansion as discrete curvature. Spectral/coboundary expansion (Ramanujan com-

plexes) stabilizes decoding and mirrors negative curvature/convexity that underlies unique-
ness of extremals.

Dependency sketch (who supplies what).

• Extremal entropy functionals: Ryu–Takayanagi → Lewkowycz–Maldacena (replica) → FLM

(bulk term) → Dong (higher curvature).

• Reconstruction mechanisms: RT/FLM/Dong → wedge reconstruction (tensor networks) →

scalable QLDPC (Panteleev–Kalachev) underpinned by expanders (Lubotzky).

• Interface calculus: GH/Israel ↔ Hadamard shape calculus; both inform “where” the varia-
tions live.

• Entropy notions: Shannon/Rényi/Tsallis/IG ↔ motivic entropy (Marcolli) and motivic Crofton/Vitushkin

(non-Archimedean kinematics).

Productive tensions and scope conditions.

• Extensivity vs. non-extensivity. Rényi/Tsallis adjust additivity and tail sensitivity;

scope: specify α/q and the operational regime (coding vs. thermodynamic vs. categorical).

• Area laws vs. corrections. Classical area (RT) vs. bulk entanglement (FLM) vs. higher-
curvature (Dong); scope: Einstein vs. general L(g, R, . . .), 1/N counting, renormalization
scheme for counterterms.

16
 • Existence vs. efficiency. Holographic reconstruction can exist while efficient decoders may
not; complexity bounds (Harlow–Hayden) set practical limits. Scope: specify decoder class
and resources.

• Continuum vs. discrete. Smooth extremal surfaces vs. graph cuts; approximation quality
depends on expansion, local check weight, and boundary conditions.

• Analytic vs. non-Archimedean kinematics. Euclidean Crofton vs. motivic Crofton;

scope: definability, choice of Grothendieck ring/measure, and compatibility with λ-structures.

Knobs & toggles (crosswalk).

• Gravity. Lagrangian (Einstein/higher curvature); matter content; region topology; ensem-

ble/boundary conditions; turn on/off Sbulk .

• Information geometry. Divergence family (KL vs. α); connection parameter α; pure vs.
mixed; measurement model (POVM).

• Codes/expanders. Degree and spectral gap; check weight; code rate vs. distance; decoder
class; noise model (erasure/Pauli/mixed).

• Shape calculus. Boundary conditions; coefficient jumps; geometric regularizers (perime-

ter/curvature); topological derivatives on/off.

• Motives. Choice of coefficients; topology (Nisnevich/étale); transfers on/off; exponentiable

measure; λ/Witt functoriality; valuation on K0 .

Reusable definitions and lemmas (to standardize later).

• Interface object: a triple (I, J , B) where I is the geometric support (e.g., ∂Ω, γA , closed
immersion i), J is the jump/trace law (Israel, transmission), and B the boundary contribution
to the variational functional (GH, Hadamard density, localization triangle).

• Generalized entropy functional: Sgen [Σ; ·] = Area/(4GN )+Sbulk +(higher-curvature), extrem-

ized over admissible Σ (QES). Discrete analog: minimal cut with penalties/weights reflecting
corrections.

• Adjoint reconstruction schema: a pair of maps (push, pull) (e.g., f∗ , f ∗ ; encoder/decoder;

e/m-projection) with a compatibility (triangle/orthogonality) identity specifying correctness
and stability.

What transfers where (operational recipes).

1. From metrics to areas. Use local QFI densities as sensitivity weights for shape flows of
γA ; compare stationary conditions with minimal/QES equations.

17
 2. From cuts to codes. Replace continuum regions by expander-supported Tanner complexes;
evaluate minimal-cut proxies for S(A) and compare to code distance/threshold.

3. From interfaces to motives. Translate boundary problems to localization triangles; track

“information loss” along f ∗ , f∗ , f! , f ! using motivic entropy.

4. From reciprocity to duality. Map Onsager reciprocity to (∇, ∇∗ ) and (f! , f ! ) adjunctions;
use it to design symmetric update rules in variational problems.

Limitations and risks (to keep explicit). Replica analytic continuation is an assumption; QES
localization needs control of backreaction. Coding models idealize geometry and may not admit
efficient decoders despite theoretical recoverability. Shape derivatives require regular boundaries or
enriched tools for topology change. Motivic constructions depend on definability, choice of measure,
and coefficient fields.

Bottom line. The same structural stencils—interface primacy, variational extremality, met-
ric/entropy dictionaries, expansion-driven robustness, and adjoint reconstruction—govern the grav-
ity, information, PDE, coding, and motivic worlds. Making these correspondences explicit yields
portable definitions and testable predictions that we leverage in Section 8.

8 Outlook and Open Problems

Programmatic questions.

1. Metric ↔ area dictionary. Can one derive a principled map from local sensitivity (Fisher/Bures
densities) on boundary data to first-variation weights for extremal surfaces, so that station-
arity of Sgen corresponds to vanishing “information flux” along γA ?

2. Complexity-aware variational principles. Incorporate decoding/algorithmic resources

into the gravitational variational problem: optimize Sgen subject to a constraint on the com-
putational complexity of reconstruction, and identify the induced shift of QES.

3. Discrete ↔ continuum via expansion. Quantify when expander-supported tensor net-

works approximate continuum entanglement: give explicit bounds relating spectral/coboundary
gaps to errors in areas vs. cuts and to code distance/threshold.

4. Motivic entropy flows on six functors. Formulate conservation/production laws for

motivic entropy under (f ∗ , f∗ ; f! , f ! ), identify “interface” objects (boundary motives) carrying
the loss, and classify when entropy is invariant under localization triangles.

5. Non-Archimedean kinematics of complexity. Develop Crofton/Vitushkin-inspired in-

variants as arithmetic analogs of area and cut capacity; compare their extremals with cate-
gorical reconstruction and with IG divergences on definable families.

18
 6. Shape calculus for entanglement functionals. Derive Hadamard-type formulas for the
first variation of Sgen under smooth deformations of γA (including counterterms and higher-
curvature corrections), and relate the boundary density to bulk stress tensors and QFI-like
objects on ∂A.

Toggle-style falsifiers and tests.

• Gravity (QES vs. area). Toggle: turn Sbulk on/off in solvable models (e.g., double-trace
deformations in AdS3 /CFT2 ). Prediction: QES shifts by δγA with first-order change δSgen =
γA ( 4GN K + sbulk ) Vn ds. Falsifier: violation of strong subadditivity or entanglement-wedge
R 1

nesting when bulk term is included.

• Higher curvature. Toggle: add a small R2 coupling. Prediction: entropy picks up Wald/extrinsic
terms; minimal surfaces cease to extremize the correct functional. Falsifier: disagreement with
linearized Dong functional in perturbative holographic states.

• Codes/expanders. Toggle: vary spectral gap of underlying complexes at fixed degree.

Prediction: code distance and threshold grow with gap; minimal-cut entropy approaches
continuum value within an error controlled by expansion. Falsifier: counterexamples with
large gap but poor distance/threshold.

• IG divergences. Toggle: replace KL by α-divergences in inference tasks mirroring boundary

reconstructions. Prediction: geodesic/projection changes match altered stability/robustness,
paralleling higher-derivative or bulk-entropy corrections. Falsifier: no measurable change in
sensitivity/curvature despite divergence swap.

• Shape calculus. Toggle: activate/deactivate internal interfaces Γ in PDE models approx-

imating wedges. Prediction: boundary-only gradients split across ∂Ω and Γ with Israel-like
transmission terms. Falsifier: failure of jump conditions in the shape derivative.

• Motivic side. Toggle: push a motivic measure through f ∗ , f∗ , f! , f ! on a stratified mor-

phism. Prediction: motivic entropy respects a balance law with loss localized on the boundary
stratum; Crofton/Vitushkin invariants are invariant under definable homotopies. Falsifier:
entropy increases under pullback or violates expected additivity on localization triangles.

Roadmap (short-to-long horizon).

1. Near-term. (i) Prove a first-variation identity for Sgen mirroring Hadamard form; (ii) derive
cut/area error bounds on expanderized tensor networks; (iii) specify a minimal axiomatization
of motivic entropy compatible with six operations.

2. Mid-term. (i) Construct complexity-constrained variational problems and analyze QES

shifts; (ii) implement α-divergence IG flows as priors for boundary reconstructions and com-

19
 pare with QES predictions; (iii) compute motivic Crofton/Vitushkin invariants on basic fam-
ilies (curves, tori) and test functoriality.

3. Long-term. (i) A categorical “boundary motive” capturing interface data and loss; (ii)
a rigorous discrete-to-continuum theorem linking expansion to holographic entropy; (iii) a
unified metric/entropy dictionary spanning analytic, combinatorial, and motivic regimes.

Scope conditions and risks. Replica analytic continuation and backreaction control limit grav-
itational statements; efficient decoders may be unavailable even when reconstruction exists; shape
derivatives require regularity or augmented topological tools; motivic constructions depend on de-
finability and coefficient choices. We therefore present predictions as conditional on the stated
knobs and toggles, with tests designed to fail fast if assumptions are violated.

Bottom line. By treating interfaces as carriers of law, variational principles as selectors of geom-
etry, metrics/divergences as sensitivity functionals, expansion as discrete curvature, and six-functor
calculus as categorical gluing, we obtain a coherent program for entropy across gravity, information,
computation, and motives. The tests above are intended to render the program falsifiable and to
guide concrete calculations in the sections to follow.

9 Section→Reference Concordance
§1 — Introduction
• Rényi (measures of entropy)

• Tsallis (nonextensive entropy)

• de Groot & Mazur (near-equilibrium thermodynamics)

• Amari & Nagaoka (methods of information geometry)

• Provost & Vallée (Riemannian structure of quantum states)

• Braunstein & Caves (statistical distance/quantum Fisher)

• Cartier (historical panorama: from sets to sheaves/topoi to motives)

§2 — Information Geometry and Quantum Statistical Distance

• Amari & Nagaoka

• Provost & Vallée

• Braunstein & Caves

• Rényi

20
 • Tsallis

§3 — Gravitational Entropy, Interfaces, Variational Structure

• Gibbons & Hawking (Euclidean action & boundary term)

• Israel (junctions & thin shells)

• Ryu & Takayanagi (holographic area law)

• Lewkowycz & Maldacena (replica derivation)

• Dong (higher-derivative gravity entropy functional)

• Harlow & Hayden (complexity vs. decoding/firewalls)

§4 — Entanglement, Quantum Corrections, and Error Correction

• Ryu & Takayanagi

• Lewkowycz & Maldacena

• Faulkner, Lewkowycz & Maldacena (bulk entanglement/one-loop)

• Dong (higher-derivative functional)

• Pastawski, Yoshida, Harlow & Preskill (holographic QECC)

• Panteleev & Kalachev (quantum LDPC with near-linear distance)

• Lubotzky (Ramanujan complexes & high-dimensional expanders)

§5 — Arithmetic & Categorical Backbones

• 5.0 Primer — Voevodsky’s Nordfjordeid Lectures (motivic homotopy)

• 5.1 — Cisinski & Déglise (triangulated motives, six functors)

• 5.2 — Marcolli (Motivic Information: entropy, information loss, Witt/λ)

• 5.3 — Motivic Cauchy–Crofton & Vitushkin invariants (non-Archimedean)

• Ben-Bassat & Kremnizer (relative analytic geometry, Banach settings)

§6 — Shape Optimization, Interfaces, PDE Variations

• Sokołowski & Zolesio (shape sensitivity analysis)

• de Groot & Mazur (transport, reciprocity, production)

• Israel (interface analogy for transmission/jumps)

21
§7 — Cross-Domain Synthesis
• Interface primacy: GH boundary, Israel junctions, Hadamard gradients, localization triangles

• Variational structure: replicas/QES; PDE shape derivatives

• Metric/entropy dictionary: Fisher/Bures; RT/FLM/Dong; motivic entropy; motivic V0

• Dualities/reconstruction: IG (∇, ∇) ; six functors; QECC/wedge reconstruction

• Expansion as discrete curvature: Ramanujan complexes; QLDPC

• Historical framing (Cartier)

§8 — Outlook and Open Problems

• Motivic entropy & boundary objects (six functors)

• Analytic ↔ non-Archimedean kinematics (Crofton/Vitushkin)

• Non-equilibrium beyond linear response (Hatano–Sasa link)

• Expanders to bulk discretizations; practical decoders

• Complexity-aware variational principles

• Framing note (Cartier’s “dream”)

22
A Glossary
Cross-cutting idioms (project-specific)
Stencil A reusable structural template (e.g., “interface primacy,” “variational extremality,” “met-
ric/entropy dictionary”) abstracted from multiple domains.

Knob A model choice that can be set to tune behavior (e.g., divergence family, higher-curvature
couplings, graph degree, coefficient ring).

Toggle A feature that can be switched on/off to test causal impact (e.g., include/exclude Sbulk ;
enable/disable expansion; add/remove internal interface).

Interface primacy The organizing idea that laws and variations localize on boundaries or junc-
tions (GH boundary term; Israel conditions; Hadamard boundary forms; localization trian-
gles).

Metric/entropy dictionary A mapping between notions of distinguishability (metrics, diver-

gences) and geometric/combinatorial functionals (areas, cuts, capacities).

Discrete curvature Heuristic for expansion/spectral gap playing the stabilizing role of negative
curvature/convexity in discrete structures.

Gravity & holography

AdS/CFT Duality between a (d+1)-dimensional anti–de Sitter gravity theory and a d-dimensional
conformal field theory on its boundary.

Euclidean action IE Wick-rotated gravitational action used to evaluate partition functions and
thermodynamics on classical saddles.
√
Gibbons–Hawking–York (GHY) term Boundary term −1
h K ensuring a well-posed
R
8πGN ∂M
Dirichlet variational problem in gravity.

Surface gravity κ Measure of gravitational acceleration at a Killing horizon; fixes the Hawking
temperature T = κ/(2π) in Euclidean regularity.

Israel junction conditions Relations linking jumps of extrinsic curvature to surface stress on a
thin shell: [Kab ] − hab [K] = 8πGN Sab .

RT formula Ryu–Takayanagi leading-order holographic entanglement entropy: S(A) = area(γA )/(4GN )

for minimal surface γA homologous to A.

Homology constraint Requirement that the bulk surface γA and boundary region A bound a
bulk region; enforces consistent cutting and subadditivity.

23
Replica trick Compute S(A) = (1 − n∂n ) log Zn |n=1 via n-fold branched coverings; in gravity,
implemented by cosmic branes/defects.

Cosmic brane Codimension-2 defect sourcing conical excess/deficit 2π(1 − 1/n) in replica geome-
tries; its extremality reproduces RT/QES conditions.

QES (Quantum Extremal Surface) Surface extremizing the generalized entropy Sgen (not just
area); anchors the entanglement wedge in semiclassical gravity.

Sgen (generalized entropy) Sgen = area(γ)/(4GN ) + Sbulk + (counterterms) + (higher curvature);

objective to extremize for QES.

Sbulk Entanglement entropy of bulk quantum fields across the Cauchy slice bounded by γ; provides
the FLM one-loop correction.

FLM correction Faulkner–Lewkowycz–Maldacena term adding Sbulk (plus renormalization coun-

terterms) to classical RT.

Wald entropy / Noether charge Black-hole entropy from Noether charge of diffeomorphism
invariance, ∼ ∂L/∂Rµνρσ εµν ερσ .

Dong functional General holographic entropy functional in higher-derivative gravity: Wald term
plus necessary extrinsic-curvature corrections fixed by replicas.

Binormal εµν Antisymmetric unit two-form normal to a codimension-2 surface; contracts the
Riemann derivative in Wald-like formulas.

Backreaction Change in background geometry due to matter/brane sources; must be controlled

when adding Sbulk or defects.

Counterterms Local boundary terms introduced to renormalize divergences (e.g., in asymptoti-

cally AdS spacetimes or Sbulk ).

Entanglement wedge Bulk domain of dependence bounded by A and its QES; region recon-
structible from boundary subregion A in holography.

Strong subadditivity / monogamy Inequalities for quantum entropy/mutual information; in

classical RT, geometric inequalities typically enforce them.

Information geometry & entropy

Statistical manifold Parameter space Θ of distributions endowed with a Riemannian metric and
affine connections.

Fisher information metric gij = Eθ [∂i log pθ ∂j log pθ ]; canonical metric of local distinguishabil-
ity.

24
Dual connections (∇, ∇) Torsion-free affinities dual w.r.t. g; encode mixture/exponential geome-
tries and projection theorems.
(−α)
α-connections One-parameter family ∇(α) with (∇(α) )=∇ ; α = ±1 give m/e-flatness.

e/m-projections Orthogonal projections in Bregman/KL geometry onto exponential (e) or mix-

ture (m) families; underpin iterative scaling/EM.

Bregman divergence Dφ (x∥y) = φ(x) − φ(y) − ⟨∇φ(y), x − y⟩; induces generalized Pythagorean
identities.

Fubini–Study metric Natural metric on projective Hilbert space for pure states: ds2 = ⟨dψ| dψ⟩−
|⟨ψ| dψ⟩|2 .
Q
Quantum Fisher information (SLD) gij = 1
2 Tr(ρ{Li , Lj }), Li solving ∂i ρ = 2 (Li ρ
1
+ ρLi );
bounds quantum estimation.

Bures/Helstrom metric Monotone quantum metric induced by Uhlmann fidelity; equivalent to

SLD QFI metric.

Rényi entropy Hα = 1
log p ; tunes tail sensitivity and multifractal emphasis.
P α
1−α
P
1− pq
Tsallis entropy Sq = q−1 ; nonextensive entropy modeling long-range/constraint-driven sys-
tems.

Entropy production Near-equilibrium quadratic form in forces/fluxes; nonnegative under Onsager–

Casimir reciprocity.

Onsager–Casimir reciprocity Symmetry/antisymmetry of transport coefficients under time re-

versal; Lij = ±Lji .

Codes, networks, & expanders

Tensor network Factorization of states via tensors on a graph; minimal cuts approximate entan-
glement of boundary partitions.

Perfect tensor Highly entangling tensor that realizes isometries for any bipartition up to half;
building block for holographic codes.

Operator pushing Identity allowing bulk operators to be represented on different boundary re-
gions within the entanglement wedge (complementary recovery).

Bulk→boundary isometry Encoding map V : Hbulk → Hbdy of a holographic code/QECC.

LDPC code (Quantum) low-density parity-check code with sparse parity checks (bounded weight
per stabilizer).

25
Quantum LDPC (QLDPC) Stabilizer code with sparse X/Z-checks; recent families achieve al-
most linear minimum distance in blocklength.

Distance d Minimum weight of a logical operator; number of correctable errors ⌊(d − 1)/2⌋.

Rate R Ratio k/n of logical to physical qubits/bits.

Check weight Maximum number of variables touched by a single parity check (locality of con-
straints).

Tanner graph / chain complex Bipartite graph (or higher-dimensional cellular complex) en-
coding checks vs. variables for LDPC/QLDPC.

Expander graph Sparse graph with large spectral gap; ensures rapid mixing and robust connec-
tivity.

Ramanujan complex High-dimensional expander built as quotient of Bruhat–Tits buildings with

optimal spectral properties.

Spectral gap λ Difference between trivial and next eigenvalue; controls mixing and expansion
quality.

Coboundary expansion High-dimensional generalization of edge expansion using cochains/coboundaries;

relevant to QLDPC construction.

Decoder Algorithm reconstructing logical information from noisy/broken boundary data (era-
sure/Pauli/mixed noise).

Shape calculus, PDE, & interfaces

Speed method Parametric deformation Ωt = (id + tV )(Ω) driven by a velocity field V .

Material derivative u̇ Lagrangian time derivative of the state along particle paths during defor-
mation.

Shape derivative dJ(Ω)[V ]: first variation of a functional under boundary deformation, measur-
ing sensitivity to geometry.

Hadamard structure theorem Shape derivative admits the boundary representation

Z
dJ(Ω)[V ] = G Vn ds,
∂Ω

so only the normal component of the perturbation V matters to first order.

Adjoint method A practical way to compute the shape gradient G is to introduce an adjoint
variable p solving a related PDE; this avoids explicit differentiation of the state u with respect
to domain variations.

26
Transmission (jump) conditions Continuity of state and flux across an internal interface Γ
(e.g., JuK = 0, Jκ∂n uK = 0).

Curvature regularization Perimeter/mean-curvature penalties added to stabilize ill-posed shape

Hessians.

Categorical & motivic

Sm/k Category of smooth schemes over a field k (with a Grothendieck topology like Nisnevich or
étale).

A1 -homotopy Equivalence relation contracting along the affine line; core to motivic homotopy
theory.

Sheaves with transfers Sheaves augmented with finite correspondences; enhance functoriality
beyond morphisms.

SH(k) Stable motivic homotopy category obtained by stabilizing A1 -homotopy via suspensions.

DM(X) Triangulated category of motives over a scheme X; supports six-functor formalism.

Six functors (f ∗ , f∗ , f! , f ! , − ⊗ −, Hom) with adjunctions and base-change; categorical gluing and
duality toolkit.

Localization triangle Distinguished triangles j! j ! → id → i∗ i∗ (and dually), modeling cut–paste

along a closed/open pair.

Purity Identification of i! with i∗ (up to twists/shifts) for regular immersions; categorical “bound-
ary condition” principle.

Realization functor Functor from motives to classical cohomology (Betti, ℓ-adic, de Rham) com-
patible with six operations.

Grothendieck ring K0 (Vark ) Free abelian group on isomorphism classes of varieties modulo scis-
sor relations [X] = [Z] + [X \ Z] with product by fiber product.

Motivic measure µ Ring homomorphism K0 (Vark ) → R (target ring R) assigning “sizes” con-
sistent with cut–paste and product.

Exponentiable measure Motivic measure compatible with symmetric powers (via λ-ring/Witt
structures), enabling zeta-type exponentiation.

Witt/λ-ring formalism Algebraic framework encoding symmetric powers/exterior powers; un-

derlies exponentiation of motivic measures.

Motivic entropy Information functional built from exponentiable motivic measures; behaves
functorially under changes of coefficients and morphisms.

27
Information loss (motivic) Decrease of motivic entropy along (i) ring maps R → R′ or (ii)
geometric morphisms f : X → Y ; categorical analog of coarse-graining.

Adjunction Pair of functors (L, R) with Hom(Lx, y) ∼ = Hom(x, Ry); appears as (f! , f ! ) and
(f ∗ , f∗ ), mirroring reconstruction/decoding dualities.

Boundary motive (heuristic) Prospective object capturing interface contributions within DM(X)
analogous to GH/Israel/Hadamard boundary terms.

Non-Archimedean & integral geometry

Berkovich space Analytification of algebraic varieties over non-Archimedean fields; locally con-
tractible, path-connected spaces suitable for analysis.

Non-Archimedean Cauchy–Crofton Integral-geometric formulas counting incidences in defin-

able families with values in motivic/constructible measures, generalizing Euclidean Crofton.

Vitushkin invariants Vk Measures of geometric complexity (covering/oscillation) of sets; motivi-

cally adapted versions quantify arithmetic complexity/“entropy.”

Valuation on K0 Map assigning sizes to classes in the Grothendieck ring, often respecting mul-
tiplicative/additive structures; used to evaluate motivic integrals/invariants.

Complexity & limits

Complexity barrier (Harlow–Hayden) Hypothesis that decoding Hawking radiation to ex-
tract detailed interior information requires resources exponential in black-hole entropy; sep-
arates existence from feasible reconstruction.

Efficient decoder Algorithm with polynomial (or otherwise bounded) resource usage; may not
exist even when information-theoretic recovery is possible.

Renormalization scheme Choice of counterterms/regulators determining finite parts of IE or

Sbulk ; scheme-dependence must be tracked across comparisons.

28
B Acronym Index
AdS Anti–de Sitter (spacetime of constant negative curvature)
CFT Conformal Field Theory
AdS/CFT Anti–de Sitter / Conformal Field Theory correspondence (holographic duality)
RT Ryu–Takayanagi (holographic area law for entanglement entropy)
QES Quantum Extremal Surface (extremizes Sgen )
HEE Holographic Entanglement Entropy
EW Entanglement Wedge (bulk region reconstructible from a boundary subregion)
FLM Faulkner–Lewkowycz–Maldacena (one-loop/quantum correction to RT via Sbulk )
GHY (YGH) Gibbons–Hawking–York boundary term (also written GH or YGH)
GN Newton’s constant GN
SSA Strong Subadditivity (of von Neumann entropy)
MI / TMI Mutual Information / Tripartite Mutual Information
QEC Quantum Error Correction (general notion)
QECC Quantum Error-Correcting Code (a specific code)
LDPC Low-Density Parity-Check (class of sparse codes)
QLDPC Quantum Low-Density Parity-Check (quantum analog of LDPC)
POVM Positive Operator-Valued Measure (general quantum measurement)
IG Information Geometry
KL Kullback–Leibler (divergence)
QFI Quantum Fisher Information (metric on quantum statistical models)
SLD Symmetric Logarithmic Derivative (defines QFI via ∂i ρ = 21 (Li ρ + ρLi ))
FS Fubini–Study (metric on projective Hilbert space)
CRB / QCRB (Quantum) Cramér–Rao Bound (estimation lower bound)
EM Expectation–Maximization (projection-based inference algorithm)
PDE Partial Differential Equation
DM Triangulated category of motives DM(X) (Cisinski–Déglise framework)
SH(k) Stable Motivic Homotopy Category over a field k
A1 Affine line; A1 -homotopy denotes contraction along the affine line
Sm/k Category of smooth k-schemes (with a Grothendieck topology)
Gm Multiplicative group scheme
K0 (Vark ) Grothendieck ring of varieties over k
λ-ring / Witt λ-ring / Witt-vector formalism (symmetric/exterior power operations)
NA Non-Archimedean (e.g., Berkovich analytic geometry)
CV (Here) Crofton–Vitushkin (integral geometry and complexity invariants)
V0 Zeroth Vitushkin invariant (complexity/“entropy”-like measure)
DoF Degrees of Freedom (context-dependent)
b.c. Boundary conditions (Dirichlet/Neumann/Robin, etc.)
GH Gibbons–Hawking (often used for GHY term shorthand)

29
C Symbol Index
Conventions
• Greek indices µ, ν, ρ, σ run over spacetime coordinates; Latin indices a, b run along hypersur-
faces; i, j, k often index parameters or components in information geometry.

• Boldface or calligraphic letters denote operators, functionals, or densities (e.g., G, K).

• dx, ds, dA denote volume/area/length elements; ∂n is the derivative in the outward normal
direction n.

• Context may reuse symbols with different meanings (e.g., κ as conductivity vs. surface gravity;
λ as spectral gap, or as λ-ring operations). Disambiguation is by section.

General & basic notation

R, C, N Real, complex, and natural numbers.

id Identity map (context-dependent: category-theoretic or geometric).

Tr Trace of a linear operator/matrix.

vol(·), area(·) Volume and area functionals.

⟨·, ·⟩ Inner product; in quantum contexts Dirac brackets ⟨ψ|ϕ⟩.

d Differential/measure symbol (e.g., dd x).

∇ Gradient/covariant derivative; ∇τ is tangential divergence on a boundary.

∂Ω Boundary of a domain Ω; Γ denotes an internal interface.

JXK Jump of X across an interface (value on + side minus value on − side).

Gravity, holography, and geometry

AdS, CFT Anti–de Sitter spacetime and conformal field theory.

GN Newton’s constant GN .

IE Euclidean (on–shell) gravitational action.

R, Rµνρσ Ricci scalar and Riemann curvature tensor.

Λ Cosmological constant.

hµν Induced metric on a hypersurface; h ≡ det hab (in adapted coordinates).

Kab , K Extrinsic curvature of a hypersurface and its trace K = hab Kab .

30
nµ Unit normal vector to a hypersurface (outward-pointing unless stated).

κ (Gravity) Surface gravity of a Killing horizon; T = κ/(2π).

Z Partition function (path integral), Z ∼ e−IE in the saddle approximation.

β Inverse temperature, β = 1/T .

γA Bulk codimension-2 minimal (or extremal) surface homologous to boundary region A.

SA Entanglement entropy of boundary region A.

ΣA Bulk Cauchy slice bounded by A and γA .

Sbulk (ΣA ) Bulk matter entanglement across ΣA .

area(γ)
Sgen Generalized entropy: Sgen = + Sbulk + (counterterms) + (higher-curvature).
4GN
εµν Binormal two-form to a codimension-2 surface (normalized).

L Lagrangian density of the gravitational theory (context: higher-derivative gravity).

K Mean curvature (trace of extrinsic curvature) of an entangling surface; appears in first-variation

of area.

QES Quantum extremal surface: extremizer of Sgen .

Replica method and corrections

n Replica index in n-fold branched covers.

Zn Partition function on the n-sheeted replica geometry.

∂n Derivative w.r.t. replica index n in (1 − n∂n ) log Zn |n=1 .

Cosmic brane Codimension-2 defect sourcing the conical excess 2π(1 − 1/n).

Wald entropy Noether-charge entropy ∼ ∂Rµνρσ εµν ερσ .

∂L

Dong functional Wald term plus extrinsic-curvature corrections for higher-derivative gravity en-
tanglement.

Information geometry and statistical distance

Θ Parameter manifold for a statistical model.

pθ (x) Probability density/mass function parametrized by θ ∈ Θ.

∂i Partial derivative w.r.t. parameter θi .

31
Eθ [· ]Expectationwithrespecttopθ .

gij (θ) Fisher information metric Eθ [∂i log pθ ∂j log pθ ].

∇, ∇ Dual torsion-free affine connections on Θ.

(−α)
∇(α) α-connection family with (∇(α) )=∇ .

φ, φ Legendre-dual convex potentials generating Bregman divergences.

Dφ (·∥·) Bregman divergence: φ(x) − φ(y) − ⟨∇φ(y), x − y⟩.

e/m-projections Exponential/mixture projections minimizing Dφ under affine constraints.

α, q Parameters in Rényi (α) and Tsallis (q) entropy families.

Quantum information geometry

H Hilbert space; P(H) its projective space.

|ψ⟩ Pure state (unit vector); ρ mixed state (density operator).

FS metric Fubini–Study line element ds2 = ⟨dψ| dψ⟩ − |⟨ψ| dψ⟩|2 .

Li Symmetric logarithmic derivative (SLD) solving ∂i ρ = 21 (Li ρ + ρLi ).

Q
gij Quantum Fisher information metric: 1
2 Tr(ρ{Li , Lj }).

dBures Bures (Helstrom) distance induced by g Q .

Tensor networks, codes, and expanders

V Bulk-to-boundary encoding isometry in a holographic code.

n, k, d Blocklength, number of logical qubits, and minimum distance of a code.

R Code rate k/n.

Check weight Maximal number of qubits/bits touched by a single parity check.

λ Spectral gap (graph/complex expansion parameter).

Cut Minimal number of network edges crossing a partition; proxy for entropy in ideal networks.

32
 Shape calculus, PDEs, and interfaces
Ω Spatial domain; ∂Ω its boundary.

V (Shape calculus) Deformation velocity field; Vn = V · n normal speed.

Ωt Deformed domain (id + tV )(Ω).

u, p State and adjoint fields (solutions of PDE and its adjoint).

J(Ω) Objective functional depending on geometry/state.

dJ(Ω)[V ]Shapederivative(f irstvariation)indirectionV.Hadamardshapegradientdensityon∂Ω (and

on internal interfaces Γ).

Mean curvature of ∂Ω (or of a surface).

H G

∂n Normal derivative, ∂n u = ∇u · n.

κ (PDE) Material coefficient (e.g., conductivity/diffusivity).

f Forcing term in PDEs; ud desired state.

Γ Internal material interface (across which coefficients may jump).

Per(Ω) Perimeter (surface measure) of ∂Ω.

u̇ Material derivative of u along the deformation flow.

Motives, six functors, and motivic measures

Sm/k Category of smooth k-schemes.

A1 Affine line; A1 -homotopy equivalences contract along A1 .

SH(k) Stable motivic homotopy category over k.

DM(X) Triangulated category of motives over a scheme X.

f ∗ , f∗ , f! , f ! Six-functor operations (pullback, pushforward, exceptional pushforward/pullback).

⊗, Hom Monoidal product and internal Hom in DM.

i : Z ,→ X, j : U ,→ X Closed/open immersions defining a localization pair Z ∪ U = X.

K0 (Vark ) Grothendieck ring of varieties.

µ : K0 (Vark ) → R Motivic measure with values in a commutative ring R.

λ-operations Symmetric/exterior power operations (Witt/λ-ring formalism).

ν Valuation (on K0 or a target ring) used to evaluate motivic integrals/invariants.

V0 Zeroth (motivic) Vitushkin invariant—complexity/“entropy”-like measure.

33
Thermodynamics and transport
Ji Fluxes (e.g., heat, mass) in linear irreversible thermodynamics.

Xj Thermodynamic forces (affinities, gradients).

Lij Transport coefficients with Onsager–Casimir reciprocity Lij = ±Lji .

σ Entropy production rate (typically quadratic in X near equilibrium).

Miscellaneous
δ Variation operator (first-order change of a functional or surface).

β Inverse temperature; also used for exponents in some contexts (disambiguated locally).

◦ Function composition.

∼
=, ≃ Isomorphism/homotopy equivalence (context-dependent).

D Experiment Matrix & Falsifiers

E1 — Gravity (QES vs. area) — §3, §4
Knob/Toggle: Turn Sbulk on/off (same state/region). Z
Prediction: QES shifts by δγA ; first variation obeys δSgen = 1
+ sbulk Vn ds; wedge


4GN K
γA
nesting preserved.
Falsifier: Strong subadditivity or entanglement-wedge nesting violated after including Sbulk ; sta-
tionary condition fails to match stress/QES equations.
Status/Result: Proposed; not yet performed.

E2 — Higher-curvature corrections — §3, §4

Knob/Toggle: Add small R2 (or GB-like) coupling.
Prediction: Entropy functional gains Wald + extrinsic terms; extremal surface moves off minimal-
area locus per Dong (linear response).
Falsifier: Measured extremality/entropy disagrees with linearized Dong prediction; regulator-
scheme inconsistencies.
Status/Result: Proposed; not yet performed.

E3 — Discrete ↔ continuum (tensor networks vs. geometry) — §4, §7

Knob/Toggle: Replace continuum region by expander-supported TN; vary spectral gap λ.
Prediction: Minimal-cut entropy approaches continuum S(A) with error decreasing in λ; better
expansion ⇒ larger code distance/threshold.

34
Falsifier: Large λ but poor distance/threshold; cut entropy deviates from S(A) beyond stated
bound as λ increases.
Status/Result: Proposed; not yet performed.

E4 — Coding robustness on expanders — §4

Knob/Toggle: Hold degree/check weight fixed; sweep λ.
Prediction: Logical error rate decreases; distance d grows (near-linearly for best families); decod-
ing succeeds under stated noise.
Falsifier: No monotone improvement of distance/threshold with increasing λ under fixed locality
constraints.
Status/Result: Proposed; not yet performed.

E5 — IG divergence swap — §2, §7

Knob/Toggle: Replace KL by α-divergences in matched inference tasks.
Prediction: Geodesics/projections and curvature (sensitivity) shift; robustness–sensitivity trade-
off tracks α.
Falsifier: No measurable change in curvature/sensitivity or projection behavior across α after
controlling estimators/priors.
Status/Result: Proposed; not yet performed.

E6 — Shape-calculus interfaces — §6
Knob/Toggle: Activate/deactivate internal interface Γ; vary coefficient jumps.
Prediction: Shape derivative splits over ∂Ω and Γ with Israel-like transmission terms; GΓ matches
flux/energy jumps.
Falsifier: Failure of jump conditions in dJ despite regular solutions; tangential components affect-
ing first-order dJ (violates Hadamard structure).
Status/Result: Proposed; not yet performed.

E7 — Complexity-aware QES — §8
Knob/Toggle: Constrain decoder/algorithmic complexity in reconstruction.
Prediction: Optimal QES shifts relative to unconstrained case; monotone trade-off between Sgen
extremality and computational budget.
Falsifier: No detectable QES shift across wide complexity budgets, or non-monotone/contradictory
trade-off.
Status/Result: Proposed; not yet performed.

35
E8 — Non-equilibrium link (Hatano–Sasa vs. entanglement production) — §8
Knob/Toggle: Drive boundary state quasi-statically/steadily; compare excess/housekeeping terms.
Prediction: A geometric production functional correlates with ∆Sgen (or ∆Sbulk ) under controlled
deformations; reciprocity structures align.
Falsifier: No correlation between thermodynamic production and entanglement changes under
matched protocols; reciprocity violations beyond noise.
Status/Result: Proposed; not yet performed.

E9 — Motivic entropy flows on six functors — §5, §8

Knob/Toggle: Push µ : K0 (Vark ) → R along f ∗ , f∗ , f! , f ! for a stratified map.
Prediction: Balance law: motivic entropy invariant or decreases, with loss localized on boundary
stratum via localization triangles; functorial consistency holds.
Falsifier: Entropy increases under pullback; failure of additivity across j! j ! → id → i∗ i∗ ; loss not
localized.
Status/Result: Proposed; not yet performed.

E10 — Non-Archimedean kinematics (Crofton/Vitushkin) — §5, §7

Knob/Toggle: Compute motivic V0 under definable homotopies; vary valuations.
Prediction: Invariance under definable homotopies; V0 tracks complexity/“entropy” and respects
product/scissor relations; compatible with λ/Witt functoriality.
Falsifier: Sensitivity of V0 to trivial homotopies; violation of scissor/product laws; incompatibility
with λ-operations.
Status/Result: Proposed; not yet performed.

36
Afterword: A Friendly Map of the Territory
This paper braids together ideas that, at first glance, live in very different neighborhoods: black
holes and quantum fields, information and statistics, error-correcting codes and expander graphs,
motives and non-Archimedean geometry, shape optimization and boundary calculus. Our claim is
modest but ambitious: there are recurring structural patterns—what we call stencils—that make
these subjects talk to one another. If you are a mathematically curious reader, this afterword is a
plain-language tour of those stencils and why they matter.

Why entropy and geometry belong in the same sentence

Entropy began life as a thermodynamic bookkeeper—how many micro-arrangements match a
macro-appearance—and became an information unit: how hard it is to predict or compress. Geom-
etry, meanwhile, measures shape: lengths, areas, curvature. The surprising 21st-century discovery
is that, in certain quantum gravitational settings, the entropy of a region is literally the area of
a surface. No metaphor: a precise equation ties uncertainty on a boundary to the area of an
appropriate surface in the interior.
Two morals follow. First, “amount of information” can be treated as a geometric quantity.
Second, when we change a system a little—poke a boundary, alter a state—the right response to
track is often geometric: a surface moves; a boundary term changes. The practical upshot is that
we can borrow techniques across domains: the tools that extremize an area can also optimize an
information functional, and vice versa.

Boundaries carry laws

A theme runs through the paper: interfaces do the heavy lifting. Think of the rim of a drum, the
meniscus in a glass, or a fence line in a field. In physics, making sense of a gravitational action
requires adding a boundary term. In relativity, when two regions meet along a thin shell, the
“jump” of curvature across the shell relates directly to the shell’s physical content. In analysis and
engineering, the sensitivity of a design to small deformations often collapses to a neat integral along
the boundary—only how you push normal to the edge matters to first order.
On the algebraic side, “gluing” data across an inclusion is encoded by distinguished triangles
and adjoint functors. That might sound abstract, but it is the same idea: the difference between
inside and outside is organized at the interface. Once you tune your eye to this, the boundary
becomes less of a mere edge and more of the place where the rules crystallize.

From surfaces to cuts to codes

Another motif: minimal surfaces in the continuum become minimal cuts on a network. Soap films
taught mathematicians what “minimal” means: among all surfaces spanning a wire frame, Nature
picks the one with least area. Replace the continuous film by a web of strings and beads and the
analog is a minimal cut: the smallest number of strands you must snip to separate the web into two

37
parts. In holography, the surface’s area measures entanglement; in certain tensor-network models,
the size of a minimal cut measures a corresponding boundary entropy. This is not mere analogy;
it is a way of calculating and checking claims about quantum systems using finite combinatorial
models.
Why codes? Because a tensor network can implement an error-correcting code: it embeds “bulk”
information into “boundary” degrees of freedom in a way that resists erasures. Expanders—highly
connected but sparse graphs—make this robust and scalable. They are the discrete cousin of
negative curvature: information spreads quickly, local damage stays local, and recovery is possible
with limited overhead. When we speak of “discrete curvature,” we mean exactly this stabilizing
role of expansion.

Information geometry: turning questions into distances

Information geometry treats statistical models as curved spaces. The steeper the curvature, the
more sensitive the model is to parameter changes; the flatter, the more forgiving. Distances in this
geometry quantify distinguishability: how many samples would you need to tell two hypotheses
apart? This moves us from slogans to calculus: “hard to tell apart” becomes “short geodesic,”
“sensitive” becomes “large curvature.” Later in the paper we compare these sensitivity measures to
areas and cuts. The moral is simple: once an idea is geometric, a lot of mature machinery becomes
available—geodesics, projections, orthogonality, duality.

Motives and the algebra of gluing

Motives come from algebraic geometry, a field obsessed with how complicated shapes (varieties)
can be sliced and reassembled. The guiding vision is to build a universal bookkeeping system for
cohomological data—an “algebra of shapes” that behaves well under cutting and pasting. In our
story, motives play the role of a categorical backbone. The same formal operations that govern
gluing and restriction (pullback/pushforward and their exceptional cousins) also give a language
for tracking information: how much is lost when you pass along a map; how measures behave under
symmetric powers; how valuation-like quantities capture complexity.
If that feels abstract, keep one image: in motives, too, the interface is king. The celebrated
“localization triangle” is a formal, algebraic way of saying: the whole equals the open part plus
the closed part, and the connecting maps store the boundary data. It is our boundary calculus,
rendered in categorical ink.

Stencils, knobs, and toggles

To make cross-talk practical, we defined three workhorse notions:

• Stencils are reusable patterns: boundary primacy; variational extremality; metric/entropy

dictionaries; expansion as discrete curvature; adjoint-based reconstruction.

38
 • Knobs are dials you set: the divergence you use (KL or a one-parameter generalization), the
gravitational Lagrangian (Einstein or higher-curvature), the code’s degree and check weight,
the choice of coefficients in a motivic measure.

• Toggles are on/off switches used to test causality: add or remove a bulk-entanglement term;
switch curvature corrections on; include or drop an internal material interface; tighten or
loosen decoding complexity; turn expansion up or down at fixed locality.

This language is meant to keep us honest. If two theories are supposedly “the same in spirit,” we
should be able to align stencils, set knobs to corresponding positions, and flip comparable toggles
to see whether predictions move together.

Falsifiability as design principle

A unifying narrative is exciting only if it risks being wrong in ways we can detect. Throughout
the paper, we propose toggle-style tests. Examples: if higher-curvature terms are physically rele-
vant, then the right entropy functional is not pure area and minimal surfaces cease to be correct
extremals; if expansion really is discrete curvature, then increasing spectral gap at fixed locality
should monotonically improve code distance and threshold; if bulk entanglement genuinely shifts
the optimal surface, wedge nesting should survive the shift and first variations should balance the
change in area against the change in bulk entropy.
None of these tests requires faith in a grand synthesis; they require only that we can move a
dial and observe whether a predicted inequality, extremality condition, or reconstruction property
holds. The program stands or falls with such checks.

What about complexity?

There is a difference between what is in principle reconstructible and what is feasible for any
realistic agent. Black-hole decoding schemes that demand resources exponential in the entropy are,
for all practical purposes, a no-go—even if a theorem says recovery exists. We therefore suggest
mixing computational constraints into variational principles. The question is no longer “which
surface extremizes a given entropy?” but “which surface extremizes a constrained entropy, given a
budget of allowed algorithms?” This reframes part of physics as a resource theory: time, memory,
and precision enter the laws by limiting which solutions are physically meaningful.

Why non-Archimedean geometry appears

Finally, why venture beyond the familiar real/complex numbers? Because certain patterns—
Crofton formulas that turn lengths into incidence counts, or invariants that measure geometric
complexity—have arithmetic counterparts. In non-Archimedean settings (Berkovich spaces) these
tools become definable and live naturally inside the motivic bookkeeping. When you recast “mea-
sure” as “valuation on a ring of shapes,” entropy becomes a question about how sizes behave under

39
algebraic operations. This is not a replacement for the analytic story; it is a parallel track with
the same stencils: interfaces/gluing, variational structure (in categorical form), and metric-like
invariants.

Limits of the map

Three cautionary notes. First, replica methods in gravity involve analytic continuations that are
plausible but delicate; results can depend on how you take limits. Second, network models that
match areas with cuts are idealized; real quantum field theories bring continuum subtleties, renor-
malization, and backreaction. Third, categorical constructions depend on choices of coefficients
and topologies; conclusions live under those choices. Our aim is not to hide these constraints but
to isolate them as knobs and toggles, so that scope is explicit.

So what have we learned?

If you strip away domain-specific jargon, five messages remain.

1. Interfaces are where law happens. Whether it is a boundary term, a junction condition,
a Hadamard shape gradient, or a localization triangle, the core constraints live on the edges.

2. Extremality organizes phenomena. Entropies in gravity and costs in design are selected
by variational principles; once you know the object to extremize, you inherit a calculus.

3. Distances matter. Sensitivity and distinguishability can be turned into metrics; metrics
can be compared with areas and cuts.

4. Expansion stabilizes. In discrete worlds, spectral gap plays the role of curvature: it spreads
information and supports recovery.

5. Gluing has an algebra. Motives and six functors provide a disciplined language for “how
parts make wholes,” including how information degrades or survives through maps.

Bringing these together does not collapse physics, information, computation, and arithmetic
into a single theory. It does something more practical: it gives a shared toolkit. With stencils,
knobs, and toggles, we can translate questions, design tests, and—most importantly—decide when
two stories are genuinely the same and when they only rhyme. If the program succeeds, it will be
because the boundaries held and the experiments—mathematical or numerical—answered cleanly.
If it fails, we will know which knob we mis-set and which toggle does not commute across domains.
Either way, the method is the message.

40
Technical Epilogue
We close by packaging the paper’s correspondences into a set of interoperable formalisms. The
guiding objects are: (i) interface carriers (geometric, combinatorial, categorical), (ii) variational
functionals (analytic and discrete), and (iii) sensitivity measures (metrics, cuts, valuations). We
write down axioms, variational identities, and commuting diagrams that allow cross-domain trans-
port of statements with explicit scope.

A. Interface objects and functoriality

Definition D.1 (Interface object). An interface object is a triple (I, J , B) with support I (a
codimension-1 or -2 locus: ∂Ω, γA , an immersion i : Z ,→ X, or a cellular subcomplex), a jump/trace
law J (e.g. Israel [Kab ], PDE transmission J·K, or boundary triangle j! j ! → id → i∗ i∗ ), and a
boundary contribution B to a variational functional (GHY term, Hadamard density, connecting
morphism).

Definition D.2 (Reconstruction adjunction). A reconstruction adjunction is a pair (push, pull)

with a triangular identity ensuring stable recovery: holographically (bulk operator support ⊂
EW(A)) we have pullA ◦ pushA ≃ id on the admissible subalgebra; in motives (f! , f ! ) and (f ∗ , f∗ )
satisfy analogous identities; in information geometry (IG) the e/m-projection orthogonality is the
convex-analytic analog.

These two definitions encode §7’s stencils as morphisms in a common diagrammatic language.

B. Variational bicomplex for entropy functionals

Let S denote admissible surfaces/cuts (analytic: embedded codim-2; discrete: edge sets of cellu-
lar complexes; categorical: closed immersions i in a localization pair). Let D denote admissible
divergences/valuations (IG divergences; Sbulk densities; motivic valuations).

Analytic sector (gravity + PDE). For a deformation vector field V and surface embedding
ι : Σ ,→ M , write the first variation of generalized entropy
Z  
δSgen [Σ; V ] = 1
4GN K + sbulk + sh.d. Vn dA + (corner/ct) , (12)
Σ

with K the mean-curvature trace, sbulk the modular-energy density for bulk fields on the entan-
glement cut, and sh.d. the higher-derivative correction density (extrinsic/Wald-like). The QES
condition is δSgen = 0 for all admissible normal V (§4). The Hadamard formula (§6) is the PDE
analog: Z Z
δJ(Ω)[V ] = G Vn ds + GΓ Vn ds , (13)
∂Ω Γ

with Γ internal interfaces and GΓ a transmission-density (Israel analog).

41
Discrete sector (networks + codes). For a tensor network G and boundary cut C ⊂ E(G),
define
Sdisc (A) = min δSdisc = (14)
X X
we , δwe ,
C∈Sdisc (A)
e∈C e∈C ∗

where weights we encode local bond dimensions/corrections; variation δwe captures code or geome-
try perturbations. Expansion parameters (spectral/coboundary gaps) control admissible deforma-
tions and guarantee stability of C ∗ (§4).

i j
Categorical/arithmetic sector (motives). For a localization pair Z ,→ X ←- U and exponen-
tiable motivic measure µ : K0 (Vark ) → R, define a motivic “boundary balance”:

∆∂ Sµ (X; Z) := Sµ (X) − Sµ (U ) − Sµ (Z) = Lµ (i, j) , (15)

where Sµ is a (logarithmic) entropy functional derived from µ and Lµ is the localized loss carried
by the boundary triangle (Appendix 9, §5). In non-Archimedean Crofton/Vitushkin calculus, V0
plays the role of a complexity/entropy valuation stabilized under definable homotopies.

C. Metric–area/valuation dictionary (axioms)

We propose an axiomatization ensuring consistency across sectors.

A1 (Monotonicity). For admissible inclusions A ⊆ B (analytic: nested regions; discrete: nested

vertex sets; categorical: U ⊂ X), the sensitivity/entropy functional is monotone: S(A) ≤
S(B); motivically, Sµ (U ) ≤ Sµ (X) after fixing a valuation/normalization.

A2 (Subadditivity/SSA stencil). For overlapping regions (A, B), the functional obeys S(A) +
S(B) ≥ S(A∪B)+S(A∩B) (or its discrete/categorical analog via Mayer–Vietoris/localization).
Violations signal scope failure (e.g. regulator choices).

A3 (Stability under coarse graining). Under admissible coarse-graining C (RG, decoder pro-
jection, or change of coefficients R → R′ ), S does not increase: S ◦ C ≤ S; the deficit is
loss.

A4 (Infinitesimal stationarity). Extremals satisfy a normal-variation Euler condition: (12) or

(13) equals 0 for all admissible Vn ; discrete analog uses cut stability under local flips; motivic
analog: vanishing boundary loss (15).

A5 (Functorial compatibility). For composable maps, the entropy/loss is compatible with ad-
junction triangles (IG orthogonality; categorical base-change; code encoder/decoder dia-
grams).

Assuming A1–A5, the crosswalks in §7 become diagrammatically functorial rather than analogical.

42
D. Complexity-aware extremization
We incorporate computational cost C as a penalty to form a constrained variational principle:

Jλ [Σ; A] = Sgen [Σ] + λ C(A) , δJλ = δSgen [Σ] +λ δC(A) = 0, (16)

| {z } | {z }
geometric algorithmic

where A is a reconstruction algorithm (decoder class). The Euler condition yields a Lagrange-
multiplier correction to QES data: in practice λ trades extremality against feasible recovery. Dis-
cretely, one replaces minimal cut by a penalized cut (weights absorb decoding budget). Categor-
ically, one restricts to realizations/pushforwards computable within a resource bound (e.g. con-
structible functors), with loss measured motivically. This frames §8’s E7 as a bona fide stationarity
problem.

E. Discrete-to-continuum stability via expansion

Let (Xn ) be a sequence of cell complexes with degree bounded by d and coboundary expansion
εn → ε∗ > 0, approximating a smooth negatively curved bulk. For boundary regions A ⊂ ∂Xn
and corresponding smooth regions on ∂M , assume boundary triangulations converge in Gromov–
Hausdorff sense. Then the consistency conjecture is:

(n)
Sdisc (A) − Scont (A) ≤ F (d) ε−1
n ∆n , ∆n → 0, (17)

for an explicit F (d) measuring locality constraints. Equation (17) states that expansion (discrete
curvature) suppresses the gap between minimal cuts and minimal surfaces under controlled meshing;
it organizes E3–E4.

F. Motivic balance and Crofton invariance

For exponentiable µ and valuation ν, define Sµ (X) = ν Zetaµ (X) with a fixed normalization.

Then:

Proposition D.3 (Localization balance, motivic form; heuristic). For a closed/open pair i : Z ,→
X, j : U ,→ X,
Sµ (X) = Sµ (U ) + Sµ (Z) + Lµ (i, j),

with Lµ (i, j) ≥ 0 under admissible ν, and Lµ = 0 for clean interfaces (purity/regular immersions).

In the non-Archimedean Crofton setting, for definable families F of linear data one expects
Z
V0 (X) = #(incidences with X) dµmot , V0 (X) invariant under definable homotopies,
F

putting V0 on the same axiomatic footing as analytic lengths/areas and combinatorial cut capacities.

43
G. Consistency checks across sectors
• SSA & nesting. In the analytic sector, SSA follows from minimal (or quantum extremal)
surface inequalities; in networks, from submodularity of cuts; motivically, from additivity
and positivity of Lµ . Violations isolate missing hypotheses (e.g. regulator dependence, non-
expansion).

• Orthogonality/adjunction. IG Pythagorean theorems for e/m-projections translate to

correctability triangles in QEC and to (f! , f ! )/ (f ∗ , f∗ ) adjunctions; all are instances of a
single “triangle identity” guaranteeing reconstruction.

• First variations. Equations (12)–(13) unify extremality conditions; discrete flips stabilize
cuts; motivic balance kills boundary loss. These are three faces of the same stationarity
principle.

H. Normalization and regimes

• Orders in 1/N . Sgen = O(N 2 ) + O(N 0 ) + · · · with area at leading order, bulk entanglement
at O(N 0 ), higher curvature tracking suppressed couplings.

• Locality constraints. In codes, bounded check weight is the discrete analog of derivative
order bounds in gravity; both limit admissible deformations and appear in F (d) of (17).

• Definability. Motivic statements require definable families/topologies and exponentiable

measures; violations of definability break Crofton-type invariance.

I. Closing synthesis
The paper’s claims can be reformulated as the existence of a tri-functorial schema

Analytic surfaces, δSgen ←→ Discrete cuts, δSdisc ←→ Categorical interfaces, ∆∂ Sµ ,

such that: (i) stationarity is equivalent across the three via interface objects, (ii) monotonic-
ity/SSA/orthogonality correspond to positivity/additivity/adjunction, (iii) curvature/expansion/valuation
play the role of stabilizers, and (iv) algorithmic resources enter as a penalty that deforms extremal-
ity in a controlled manner (16).
The framework is intentionally modular: replacing IG divergences, changing higher-curvature
couplings, swapping code families, or choosing different motivic coefficients corresponds to turning
knobs; adding/removing bulk terms, internal interfaces, or expander gaps corresponds to flipping
toggles. The falsifiers in Appendix D are designed to probe each module independently and then
in concert.
If the axioms (A1–A5) hold under stated regimes, the cross-domain dictionary ceases to be
metaphorical: it becomes a transport mechanism for theorems. If a falsifier fires, the failure localizes

44
to a knob/toggle and we learn whether the obstruction is analytic (e.g. backreaction), combinatorial
(e.g. non-expansion), or categorical (e.g. non-exponentiable measure). In all cases, the interface
remains the organizing locus where geometry, information, and motives trade constraints.

References
[1] S. Ryu and T. Takayanagi, “Holographic Derivation of Entanglement Entropy from
AdS/CFT,” arXiv:hep-th/0603001 (2006).

[2] A. Lewkowycz and J. Maldacena, “Generalized Gravitational Entropy,” arXiv:1304.4926

(2013).

[3] T. Faulkner, A. Lewkowycz, and J. Maldacena, “Quantum Corrections to Holographic Entan-

glement Entropy,” arXiv:1307.2892 (2013).

[4] X. Dong, “Holographic Entanglement Entropy for General Higher Derivative Gravity,”
arXiv:1310.5713 (2013).

[5] G. W. Gibbons and S. W. Hawking, “Action integrals and partition functions in quantum
gravity,” (classic ref.; see project PDF for exact edition) (1977).

[6] W. Israel, “Singular Hypersurfaces and Thin Shells in General Relativity,” Il Nuovo Cimento
B (1966).

[7] D. Harlow and P. Hayden, “Quantum Computation vs. Firewalls,” arXiv:1301.4504 (2013).

[8] F. Pastawski, B. Yoshida, D. Harlow, and J. Preskill, “Holographic quantum error-correcting

codes: Toy models for AdS/CFT,” arXiv:1503.06237 (2015).

[9] P. Panteleev and G. Kalachev, “Quantum LDPC Codes with Almost Linear Minimum Dis-
tance,” arXiv:2012.04068 (2021).

[10] A. Lubotzky, “Ramanujan Complexes and High Dimensional Expanders,” arXiv:1301.1028

(2013).

[11] (*) R. M. Wald, “Black hole entropy is Noether charge,” Phys. Rev. D 48 (1993), 3427–3431.

[12] (*) V. Iyer and R. M. Wald, “Some properties of Noether charge and a proposal for dynamical
black hole entropy,” Phys. Rev. D 50 (1994), 846–864.

[13] (*) N. Engelhardt and A. C. Wall, “Quantum Extremal Surfaces: Holographic Entanglement
Entropy beyond the Classical Regime,” JHEP (2015).

[14] S.-I. Amari and H. Nagaoka, Methods of Information Geometry, AMS & Oxford (2000).

45
[15] J. P. Provost and G. Vallée, “Riemannian structure on manifolds of quantum states,” Comm.
Math. Phys. 76 (1980), 289–301.

[16] S. L. Braunstein and C. M. Caves, “Statistical distance and the geometry of quantum states,”
Phys. Rev. Lett. 72 (1994), 3439–3443.

[17] A. Rényi, “On measures of entropy and information,” Proc. Fourth Berkeley Symp. Math. Stat.
Prob. (1961).

[18] C. Tsallis, “Possible generalization of Boltzmann–Gibbs statistics,” J. Stat. Phys. 52 (1988),

479–487.

[19] S. R. de Groot and P. Mazur, Non-Equilibrium Thermodynamics, Dover / North-Holland

(classic text).

[20] T. Hatano and S.-i. Sasa, “Steady-State Thermodynamics of Langevin Systems,” Phys. Rev.
Lett. 86 (2001), 3463–3466; arXiv:cond-mat/0010405.

[21] (*) D. Petz, “Monotone metrics on matrix spaces,” Linear Algebra Appl. 244 (1996), 81–96.

[22] (*) N. N. Chentsov, Statistical Decision Rules and Optimal Inference, Amer. Math. Soc. (1972).

[23] (*) A. Uhlmann, “The ‘transition probability’ in the state space of a ∗ -algebra,” Rep. Math.
Phys. 9 (1976), 273–279.

[24] (*) D. Bures, “An extension of Kakutani’s theorem on infinite product measures to the tensor
product of semifinite W ∗ -algebras,” Trans. Amer. Math. Soc. 135 (1969), 199–212.

[25] J. Sokołowski and J.-P. Zolesio, Introduction to Shape Optimization: Shape Sensitivity Analy-
sis, Springer (1992).

[26] D.-C. Cisinski and F. Déglise, Triangulated Categories of Mixed Motives, (see project PDF for
edition).

[27] O. Ben-Bassat and K. Kremnizer, “Non-Archimedean analytic geometry as relative algebraic

geometry,” arXiv:1312.0338 (2013).

[28] (Nordfjordeid Lectures, published notes), “A1 -Homotopy and Motivic Homotopy Theory
Primer,” (lecture notes; see project PDF).

[29] M. Marcolli, “Motivic Information,” (2018; preprint/article; see project PDF).

[30] (Authors per project file), “Non-Archimedean Cauchy–Crofton and Vitushkin-type invariants,”
arXiv:2206.15412 (v2).

[31] P. Cartier, “A mad day’s work: from Grothendieck to motives,” Bull. Amer. Math. Soc. 38
(2001), 389–408.

46
[32] (*) J. W. York, Jr., “Role of conformal three-geometry in the dynamics of gravitation,” Phys.
Rev. Lett. 28 (1972), 1082–1085.

[33] (*) V. E. Hubeny, M. Rangamani, and T. Takayanagi, “A covariant holographic entanglement

entropy proposal,” JHEP (2007).

[34] (*) A. C. Wall, “Maximin surfaces, and the strong subadditivity of the covariant holographic
entanglement entropy,” Class. Quantum Grav. 31 (2014) 225007.

[35] (*) D. Petz and C. Sudar, “Geometries of quantum states,” J. Math. Phys. (2008).

[36] (*) M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, Cam-
bridge Univ. Press (2000).

47
The 90-second briefing
This paper argues that “entropy” (how much can happen or be known) and “geometry” (how
shapes extremize) are two faces of the same coin. In gravity, a region’s quantum uncertainty is the
area of a special surface in spacetime; in information theory, distances between models measure how
distinguishable they are; in coding, minimal cuts on networks track how much information can flow.
Across all of these, the action happens at interfaces: boundary terms in gravity, junction conditions
in PDEs, cut sets in networks, and gluing triangles in modern algebra. The unifying playbook is to
extremize the right functional, read sensitivity as a metric, and use expansion (a kind of discrete
curvature) to make reconstructions robust. The paper ends with toggle-style tests—flip a knob
(add bulk entanglement, turn on higher curvature, increase expansion) and see if the predicted
extremals and inequalities still hold.

The feature-length explainer

1. Why join entropy and geometry now? Because we have precise formulas tying quantum uncer-
tainty on a boundary to the area of an interior surface, and mature tools in statistics and coding
that speak the same mathematical language. 2. Interfaces carry laws. Gravity needs boundary
terms to make variations well-posed; thin shells satisfy junction rules; shape calculus collapses
first-order changes to boundary integrals; in algebraic geometry, “localization triangles” formalize
cut-and-paste. 3. Extremality organizes the story. In holography, entanglement is set by a minimal
(or quantum extremal) surface; in PDE-constrained design, shape gradients vanish at optima; in
networks, minimal cuts bound information. 4. Distances become tools. Fisher/Bures metrics,
familiar in statistics and quantum metrology, quantify distinguishability and curvature; they can
be compared to areas and cuts. 5. Discrete curvature stabilizes. High-dimensional expanders
make error-correcting codes both sparse and powerful, echoing the role negative curvature plays
in continuum geometry. 6. Motives add a categorical backbone. The six-functor formalism turns
“glue these pieces” into functorial algebra; motivic notions of information capture loss under maps
and coefficient changes. 7. Complexity matters. Some reconstructions exist “in principle” but are
infeasible; the paper proposes adding computational budget as a penalty in the variational problem,
shifting the optimal surface/code/cut. 8. Falsify it. The work lists concrete switches—bulk en-
tanglement on/off, higher-curvature terms small/zero, expander gap up/down—and the outcomes
that must follow if the dictionary is real (e.g., nesting of entanglement wedges, entropy inequalities,
decoding thresholds).

The enthusiast’s brief

• Interface primacy: Wherever two regions meet, the rules crystallize. In gravity, the Gibbons–
Hawking–York term fixes the action; in relativity, Israel conditions tie curvature jumps to surface
stress; in shape calculus, only the normal boundary speed contributes at first order; in motives,
localization triangles encode how wholes decompose into open and closed parts. • Variational

48
stencil: A single pattern recurs—write a functional; push the boundary; set the normal variation
to zero; get an extremal locus. That is the RT/QES surface in gravity, the Hadamard gradient in
PDEs, and the minimal cut in networks. • Metric/entropy dictionary: Sensitivity metrics (Fisher,
Bures, Fubini–Study) and information divergences play the same structural role as areas and cuts;
both obey monotonicity, subadditivity, and projection/orthogonality identities. • Quantum cor-
rections and higher curvature: Beyond pure area, bulk entanglement shifts the extremal surface;
higher-derivative actions add Wald-type terms and extrinsic-curvature corrections. The framework
tracks which regimes (large-N, weak coupling) justify each term. • Codes and expanders: Holo-
graphic tensor networks instantiate bulk-to-boundary isometries with built-in erasure correction;
modern quantum LDPC codes backed by high-dimensional expanders push distance toward lin-
ear while keeping checks local—discrete negative curvature in action. • Motives and information:
Exponentiable motivic measures, λ/Witt structures, and non-Archimedean Crofton/Vitushkin in-
variants supply an arithmetic mirror of lengths/areas/cuts, with “information loss” localized by
gluing functors. • Complexity as a physical knob: Introduce a Lagrange multiplier on algorith-
mic cost; the optimal surface or cut moves when recovery must be efficient, not merely possible.
This reframes some “paradoxes” as resource mismatches. • Cross-checks and risks: The paper’s
predictions are organized into toggle-style tests. Failures localize assumptions—replica continua-
tion, regulator choices, lack of expansion, non-definable families, or computational infeasibility—so
scope is explicit rather than hand-waved. • What’s transferable: Definitions of interface objects,
generalized entropy/cost, and reconstruction adjunctions act as adapters between fields, letting
results migrate with minimal translation. • Bottom line: Treat boundaries as the carriers of law,
extremality as the selector, metrics as sensitivity, expansion as robustness, and gluing as algebra.
With those stencils, gravity, statistics, coding, PDEs, and motives become different theaters playing
variations of the same script—and we can test, not just tell, that story.

49