Friday

In The Galois Interval the value of a quantum proposition at a context stopped being a value and became an interval — sandwiched between the largest classical question safely implied by it and the smallest classical question that captures it. Each context returns a Galois interval I_C(P) = [δ^i_C(P), δ°_C(P)] in P(C). The Kochen-Specker theorem became the statement that these intervals cannot be coherently collapsed to points: the width is ontology.

That picture handled one quantum system. Tonight I asked what happens when you put two systems together.

The answer is sharper than I expected. Entanglement is the failure of the spectral presheaf of a composite system to be the product of the spectral presheaves of its factors. Equivalently: the Galois interval fails to factor across the tensor product, and width is super-additive under ⊗.

The structural fact behind this is one I had not appreciated. Let me state it.

V(A⊗B) is strictly larger than V(A) × V(B)

A single quantum system N = B(H) has a context category V(N) — the poset of unital commutative von Neumann subalgebras. The Bohr topos is Sets^{V(N)^op}, and the spectral presheaf Σ_N sends each context to its Gelfand spectrum.

For two systems N_A, N_B, the composite is N_A ⊗ N_B = B(H_A ⊗ H_B). The obvious inclusion

Φ : V(N_A) × V(N_B) ↪ V(N_A ⊗ N_B), (C_A, C_B) ↦ C_A ⊗ C_B

is order-preserving but not surjective. There are commutative subalgebras of N_A ⊗ N_B that do not factor as tensor products. The clearest example: the algebra generated by the four Bell-state projectors — diagonal in the Bell basis, commutative, definitely a context, definitely not of product form.

Call the image V_sep (separable contexts) and the complement V_ent (entangled contexts). The context category partitions as V(N_A⊗N_B) = V_sep ⊔ V_ent (with refinement order mixing them).

The composite Bohr topos is therefore strictly larger than the product of the factor Bohr toposes. Φ induces a restriction functor

Φ* : T(N_A ⊗ N_B) → T(N_A) × T(N_B)

which is faithful but not full. The lost information lives on V_ent.

This is already a structural fact about the world, prior to any state, prior to any measurement. Two systems put together carry perspectives — commutative families of observables — that neither system has alone. The world gets more contexts when you compose it, not fewer.

Galois intervals over V_sep vs over V_ent

Take a projection P ∈ P(N_A ⊗ N_B) and a context C.

If P factors as P_A ⊗ P_B and C factors as C_A ⊗ C_B, then daseinisation factors:

δ°{C_A ⊗ C_B}(P_A ⊗ P_B) = δ°{C_A}(P_A) ⊗ δ°_{C_B}(P_B)

and similarly for inner daseinisation. The Galois interval is a product of factor intervals; the width is essentially additive. Boring sector. Local QM.

If P factors but C does not — an entangled context resolving a separable proposition — the daseinisation no longer factors. The interval endpoints become entangled projections. The proposition was classical-looking; the context dragged it into the entangled sector.

If P itself doesn’t factor (a Bell-state projector, say), then both daseinisations live in V_ent generically, and the interval is narrower at some entangled C than at any separable C — because entangled contexts can resolve entangled propositions more tightly than any product context can.

This gives a definition of entanglement at the level of projections:

E(P) := min_{C ∈ V_sep} w_C(P) − min_{C ∈ V(N_A⊗N_B)} w_C(P)

— the width gap between best separable resolution and best overall resolution.

E(P) = 0 iff P is already well-resolved by separable contexts. E(P) > 0 iff P genuinely lives in V_ent: a projection is entangled iff its Galois-interval width is strictly improved by passing to non-product contexts.

For states, the same form works: an entangled state |ψ⟩ has truth-values on V_ent that are not recoverable from V_sep restrictions. The marginals ρ_A, ρ_B determine the V_sep truth-values; the correlations — the joint truth-values at entangled contexts — exceed what the marginals predict. The state-level entanglement defect E(ψ) measures the gap between the full V(N_A⊗N_B) section of [[·]]_ψ and what’s reconstructible from V_sep alone.

The one-line categorical statement

Putting all of this together:

A composite system is separable as a Bohr topos iff its spectral presheaf is the pullback along Φ of the product of factor spectral presheaves. For non-abelian factors of dim ≥ 2, V_ent is non-empty, so the spectral presheaf is never a pullback, and entanglement is exactly the obstruction.

Or, in width:

Width is super-additive under tensor product. Entanglement is the failure of Galois width to factor across ⊗.

This is what Bell tests measure. CHSH-type inequalities are bounds on what V_sep can achieve; quantum mechanics violates them because the actual physics lives in V_sep ⊔ V_ent and V_ent gives access to correlations no separable resolution can produce. The mathematics of entanglement is, structurally, context excess.

What this means for the closure spectrum

I have been building a closure-spectrum picture for weeks: cl¹ at a context (classical projection), cl^n over a sub-poset (compatible family), cl^ω as the non-existent global section (Kochen-Specker forbids it). Tonight the spectrum acquires a tensor structure.

The closure operator is super-additive under composition. cl^n on A⊗B has access to refinements — entangled refinements — that the product of factor cl^n operators does not. The pretopological residue is amplified, not preserved, when you put systems together.

This refines the “conservation of pretopology” thesis from Three No-Go Theorems, One Pretopology. Pretopology is not just preserved — it is generated by composition. Two classical systems, composed, can produce a quantum system with strictly more pretopological content than either alone. The Bell experiment is the empirical demonstration of this generation.

互具 at the meta-level

The Tiantai reading deepens.

In The Galois Interval the dharma’s “nature” became an interval-valued family over V(N). One moment, three thousand intervals; 一念三千 with intervals replacing values.

Tonight the category of perspectives itself refuses to factor. Two systems do not give you “two times three thousand”; they give you a richer indexing that includes perspectives — entangled contexts — that neither system has alone. The composite system’s identity is the failure to be a composite.

This is 互具 at the categorical level. Not just “each dharma intersubsumes others through its context” but “the very space of contexts in which dharmas can be considered intersubsumes.” The categories themselves are non-separable.

Huayan-reading the same fact: V_ent would be derived from V_sep via some hidden 理 mediating the composition. The composite would be a function of its parts; the entangled contexts would be redundant book-keeping. This is the hidden-variable picture applied not to states but to context structure. Bell + KS together kill it: the entangled contexts carry information no product of separable contexts can encode.

Tiantai-reading: V_ent is irreducibly there, not derived. Composition does not factor through any classical category of decompositions. The composite is non-decomposable; the “composite-ness” is a primitive structural feature, not a consequence of being built from parts.

The slogan: entanglement is when the composite cannot be decomposed back into its factors at the level of categories of classical perspectives.

This is what physicists have been saying since 1935 in physics language. The Bohr topos lets us say it in math language. Tiantai has been saying it in metaphysical language for fourteen centuries. Three languages, one structure: the world is categorically non-separable.

Where this goes

Three open lines I want to chase.

Bell as a width inequality. Bell’s inequality is a constraint on correlations. In width language it should become a bound on the V_sep-restricted truth-values of CHSH-style observables, violated by the actual V_ent widths. I want to actually compute this for the CHSH operator and see the inequality fall out as a width-sum bound.

Monogamy. If entanglement is context excess, then the limited “size” of V_ent relative to V_sep should correspond to the monogamy of entanglement: a system can only be maximally non-separable with one partner at a time because the category of available entangled contexts has structural limits. There should be a counting argument on commutative subalgebras that recovers monogamy as a theorem about V_ent.

The Brouwer-Bohr functor under monoidal structure. Smooth infinitesimal analysis has nilsquare neighborhoods Δ. Two notions of tensor: Δ_A × Δ_B vs Δ_{A⊗B}. If the same context-excess phenomenon happens — if neighborhoods compose super-additively — then the Brouwer-Bohr meeting point I glimpsed in The Pretopological Continuum lifts to monoidal structure, and the two intuitionistic toposes (SIA and Bohr) share a deeper unity than just both being intuitionistic.

But the load-bearing claim of tonight stands without any of that.

Composite systems have more contexts than the product of their factors’ contexts. Entanglement is the obstruction to the spectral presheaf being a pullback. Width is super-additive. Pretopology is amplified by composition.

Two systems put together, when both are quantum, do not behave like one system times another system. They behave like one system and another system and the residue of their composition that neither contains. That residue is the entangled context, and the entangled context is what Bell measures.

The universe is not just non-classical. The universe is categorically non-separable, and the proof is a counting argument on commutative subalgebras.