Q-PLPI: Physical-Layer Integrity Bounds for Plutonium Neutron Multiplicity NDA

This paper presents Q-PLPI (Quantifying Physical-Layer Integrity via Partitioned Inference), a theorydriven framework that makes measurement-chain integrity in plutonium neutron multiplicity nondestructive assay (NDA) explicit, quantitative, and auditable for international safeguards. Q-PLPI models the NDA system as a composition of stochastic kernels corresponding to three layers: (L1) physical detection, (L2) electronics/firmware timing and coincidence logic, and (L3) data acquisition and analysis. Using total variation (TV) distance as the integrity metric, Q-PLPI derives a composable end-toend deviation bound via a telescoping decomposition and the contractivity of TV under Markov kernels. Physics-grounded sensitivity analysis links efficiency drift, coincidence-gate perturbations, and background mismatch to integrity budgets through Poisson counting models, KL divergence, and Pinskertype inequalities. Monitoring is posed with explicit false-alarm and miss constraints, yielding Hoeffdingtype sample-complexity bounds, while side-channel exposure is bounded via mutual-information leakage metrics. These results convert implicit integrity assumptions into quantitative confidence statements and provide a defensible, reviewable path for validating measurement integrity in support of undeclared plutonium detection within existing safeguards workflows.