# Quantum v4.37–v4.39 integration This integration combines three checked local modules on public main `a42c09d766bb6e2f32d4c1843b9de8a71a52df6c`. The candidate chain is `acb520f3ad5e4a101b9f985a98ce450b642f0954` (v4.37), `2437a0830903a1ad673b4e8c330219b3d22ee4ad` (v4.38), and `df351face373012998d096d513c28120b9278bac` (v4.39). All three formal module files, their tests, and the registry integration are preserved byte-for-byte from the checked candidate. This integration adds shared documentation; it changes no proof, toolchain, dependency or verification rule. ## Formal coverage | Stage | Module | Named theorems | Named exports | |---|---|---:|---:| | v4.37 | QuantumNoisyBBPSSWTree | 26 | 34 | | v4.38 | QuantumScheduledNoisyRepeater | 36 | 53 | | v4.39 | QuantumNoisyRepeaterWitness | 26 | 32 | | Total | Three modules | 88 | 119 | These are code metrics, not counts of independent scientific results. The catalog contains 400 modules across 12 sections, 404 library Lean files including nested support sources and the root, and 354 direct MasterSuite imports. The registry fields are `noisy_tree`, `scheduled_noisy` and `noisy_witness`. The [v4.37 record](quantum-v4.37.md) constructs the unnormalized noisy BBPSSW tree, proves D_n = p_n W(F_n) with p_(n+1) = p_n^2 P(F_n), and gives the exact finite-depth post-swap recovery criterion. Post-depolarization preserves each node's success probability; the complete tree weight depends on the noisy orbit. The [v4.38 record](quantum-v4.38.md) composes spatial levels chronologically using List.foldl and an independently updated matrix, weight and visibility. The invariant connects these fields without recovering visibility by dividing a matrix. Raw-pair resources are the product of 2^(m+1) over the schedule, with multiplicity. Spatial segments number 2^(schedule length). The [v4.39 record](quantum-v4.39.md) certifies gamma=19/20, w0=4/5 and schedule [16,16] by rational interval chains checked in Lean. Both nonempty prefixes strictly improve the initial visibility, and the second also improves the first. Depth 16 is minimal at each of these two specific inputs. This is not global schedule optimality. Four spatial segments consume 2^34 = 17179869184 leaves in the fixed all-success resource tree. This count is neither success probability nor expected consumption with retries. The ideal [4,2] witness uses gamma=1; it is not a witness for gamma=19/20. Swapping remains ideal. Noise acts on the retained pair after each BBPSSW step. The conditional output is a normalized Werner state, generally mixed. Nonlinear diagonal resource preparation is not claimed to be a linear CP map or cloning. No asymptotic generation rate, noisy BSM model or optimal resource cost is claimed. ## Verification and publication boundaries The integration gate uses scoped/gamma quantum-physics checks against the actual public parent, including affected numeric-certificates and registry consumers. It adds a complete incremental `lake build --wfail`, a pinned independent audit of all compiled project declarations, and every local unittest module in the qualified 12 GiB / 4 CPU / swap=0 container. Live Lean, SQLite and native crypto checks are enabled. Metadata, knowledge links and whitespace are checked. Only propext, Classical.choice and Quot.sound are allowed; sorryAx and native-decide shortcuts are excluded. The integration evidence records measured results, commands, source identities, runtime limits and logs. Previous stage archives remain immutable. Candidate CI must succeed, including verify-all, before a normal fast-forward of the same SHA to main. The main push receives its own final CI receipt. Hosted profile skips are reported separately from the required zero-skip local test run. This publication creates no release tag and moves no existing tag. Leanbook v1.0.0 and its assets remain unchanged. Incremental local checks and hosted verify-all do not establish a new clean full-library beta checkpoint under VERSIONING.md. Historical development-stage records retain their original scope.