The Spectral Stability of Constraint Manifolds: Explicit Ramanujan Complexes, Operator Discrepancy, and Unconditional Separation

This volume grounds the construction in fully explicit mathematics: a bipartite Ramanujan incidence graph built via the Lubotzky–Phillips–Sarnak / Marcus–Spielman–Srivastava constructions, reduced to a bounded-degree 3-CNF formula whose satisfying assignments sit at macroscopic Hamming distance from one another. Against this variety, the book develops the Expander Cancellation Sieve: a discrepancy argument built from the Grothendieck γ₂-factorization norm and bipartite spectral expansion, proving that no signed, destructive cancellation can collapse the linear rank of the associated transfer operator below 2^Ω(n) — closing the one loophole (cancellation over ℝ or 𝔽₂) that non-negative-rank methods alone cannot touch.

From that operator-rank floor, the book derives an unconditional linear cut-width lower bound against any Boolean circuit, of any depth, over any bounded fan-in basis — a macroscopic communication bottleneck that must appear across every balanced bipartition of the variables. It further shows, unconditionally, that this floor already forces exponential size for whole classes of restricted architectures: bounded-cut-width circuits, monotone and non-cancelling circuits, and shallow (logarithmic-depth) circuits. A full chapter is devoted to confirming that the underlying sieve property evades the Razborov–Rudich Natural Proofs barrier, with an explicit, poly-time-decidable predicate and a proof that it fails the Largeness condition by a double-exponential margin — so the argument cannot be turned into a cryptographic distinguisher.

What the book does not claim is a proof that closes the remaining step for fully general circuits. The gap between a static cut-width bound and a global circuit-size bound is isolated precisely, named the Multi-Wire Amortization Obstruction, and formalized as the Cut-Capacity Accumulation Conjecture: the one open hypothesis under which the full separation NP ⊄ P/poly, and therefore P ≠ NP, would follow. The manuscript is explicit that this last step is open, not proven — the trilogy's contribution is the geometry, the unconditional partial results, and a sharply defined target for whoever closes it next, rather than a claimed resolution of the Millennium Prize problem.

Written as a direct, theorem–proof mathematical treatise (not a popular account), this volume assumes graduate-level familiarity with circuit complexity, communication complexity, and functional analysis. It will be of most interest to researchers in structural complexity, spectral graph theory, and coding theory looking for new unconditional tools — the expander-based cancellation sieve, the natural-proofs-evasion technique, and the cut-width/amortization framework are usable independently of the larger P vs. NP program they were built to serve.

This work completes The Geometry of Intractable Varieties trilogy, following Volume I (static and dynamic invariants) and Volume II (the functional-analytic and operator-theoretic bridge).