Landauer Constraints on Mesoporous Forsterite After the Locked Spin-Bath Gates

This record joins two existing lines without restating them. The silicate design papers specify Gao-type crystalline mesoporous forsterite (Mg₂SiO₄) as a durable host class and lock a Host-evaluation module on matched S0/S1 powders, with falsifiers F1–F3 still open. The Landauer–dissipation tensor and the Phonon Sink already state the thermodynamic envelope of an open quantum neural network: erasure of an unknown bit costs at least k_B T ln 2, dissipation belongs at a wall, and write-back of bath entropy into core parameters is identically zero. The manuscript places that envelope on the grain. A host can physically accommodate a multi-level encoding. A holder keeps that encoding while entropy leaves as heat rather than as logical dephasing. Native acoustic modes of the forsterite walls and the as-grown ~2.58 nm pore lattice may act as a passive sink; an engineered sink requires a specified resonator, bandgap, taper, or addressable mode. They are not interchangeable by renaming the pores. No paramagnetic density, nuclear spectral density, TLS (two-level system) spectrum, or coherence time is reported. No fidelity trajectory is claimed. Day-1 (XRD, FTIR, CW X-band EPR on matched S0/S1) remains the first experimental gate and is not executed here. If F1 or F2 fails after cleaning, Rank 5 is scaffold and encoding stays Regime C. If ¹H and pore-wall TLS are bounded and ²⁵Mg still kills the window, Branch Deplete is permitted but does not by itself make the grain a holder. If F3 fails and no nuclear window opens, Regime C is terminal on this grain. Parents: 10.5281/zenodo.21909312, 10.5281/zenodo.21923558, 10.5281/zenodo.21609954, 10.5281/zenodo.21611607, 10.5281/zenodo.20357738, 10.5281/zenodo.20597201, 10.5281/zenodo.20798938, 10.5281/zenodo.20803081, 10.5281/zenodo.22161547. Isolation and monitor records 22173068, 22206432, 22211512, and 22217506 are consumers of this constraint and are not rewritten here.